{"schema":"kbp-graph/v1","blueprint_id":"preview:b25c93e8cb43b9078ef0bf65ec4e982c3957fbc1b88be5e3f87e6af491a2e69a","lineage":"sha256:7eeaab8c36c2eb0ac97e839d20de53842404a1fdbfa176a25a2e6ab654e00987","revision":0,"exported_by":"preview: tools/preview/kbp_preview.py over `met circuit plan` (not `met blueprint export`)","arm":{"mode":"decode","rows":1,"digest":"31d4c6b3314dd12f33e3a75c6d75bca52b0d4efd27afddafef8daa37963e8ad9"},"layers":[{"index":0,"kind":"linear_attention","template":"gdn+moe_ffn","blocks":["l0.gdn","l0.moe_ffn"]},{"index":1,"kind":"linear_attention","template":"gdn+moe_ffn","blocks":["l1.gdn","l1.moe_ffn"]},{"index":2,"kind":"linear_attention","template":"gdn+moe_ffn","blocks":["l2.gdn","l2.moe_ffn"]},{"index":3,"kind":"full_attention","template":"attn+moe_ffn","blocks":["l3.attn","l3.moe_ffn"]},{"index":4,"kind":"linear_attention","template":"gdn+moe_ffn","blocks":["l4.gdn","l4.moe_ffn"]},{"index":5,"kind":"linear_attention","template":"gdn+moe_ffn","blocks":["l5.gdn","l5.moe_ffn"]},{"index":6,"kind":"linear_attention","template":"gdn+moe_ffn","blocks":["l6.gdn","l6.moe_ffn"]},{"index":7,"kind":"full_attention","template":"attn+moe_ffn","blocks":["l7.attn","l7.moe_ffn"]},{"index":8,"kind":"linear_attention","template":"gdn+moe_ffn","blocks":["l8.gdn","l8.moe_ffn"]},{"index":9,"kind":"linear_attention","template":"gdn+moe_ffn","blocks":["l9.gdn","l9.moe_ffn"]},{"index":10,"kind":"linear_attention","template":"gdn+moe_ffn","blocks":["l10.gdn","l10.moe_ffn"]},{"index":11,"kind":"full_attention","template":"attn+moe_ffn","blocks":["l11.attn","l11.moe_ffn"]},{"index":12,"kind":"linear_attention","template":"gdn+moe_ffn","blocks":["l12.gdn","l12.moe_ffn"]},{"index":13,"kind":"linear_attention","template":"gdn+moe_ffn","blocks":["l13.gdn","l13.moe_ffn"]},{"index":14,"kind":"linear_attention","template":"gdn+moe_ffn","blocks":["l14.gdn","l14.moe_ffn"]},{"index":15,"kind":"full_attention","template":"attn+moe_ffn","blocks":["l15.attn","l15.moe_ffn"]},{"index":16,"kind":"linear_attention","template":"gdn+moe_ffn","blocks":["l16.gdn","l16.moe_ffn"]},{"index":17,"kind":"linear_attention","template":"gdn+moe_ffn","blocks":["l17.gdn","l17.moe_ffn"]},{"index":18,"kind":"linear_attention","template":"gdn+moe_ffn","blocks":["l18.gdn","l18.moe_ffn"]},{"index":19,"kind":"full_attention","template":"attn+moe_ffn","blocks":["l19.attn","l19.moe_ffn"]},{"index":20,"kind":"linear_attention","template":"gdn+moe_ffn","blocks":["l20.gdn","l20.moe_ffn"]},{"index":21,"kind":"linear_attention","template":"gdn+moe_ffn","blocks":["l21.gdn","l21.moe_ffn"]},{"index":22,"kind":"linear_attention","template":"gdn+moe_ffn","blocks":["l22.gdn","l22.moe_ffn"]},{"index":23,"kind":"full_attention","template":"attn+moe_ffn","blocks":["l23.attn","l23.moe_ffn"]},{"index":24,"kind":"linear_attention","template":"gdn+moe_ffn","blocks":["l24.gdn","l24.moe_ffn"]},{"index":25,"kind":"linear_attention","template":"gdn+moe_ffn","blocks":["l25.gdn","l25.moe_ffn"]},{"index":26,"kind":"linear_attention","template":"gdn+moe_ffn","blocks":["l26.gdn","l26.moe_ffn"]},{"index":27,"kind":"full_attention","template":"attn+moe_ffn","blocks":["l27.attn","l27.moe_ffn"]},{"index":28,"kind":"linear_attention","template":"gdn+moe_ffn","blocks":["l28.gdn","l28.moe_ffn"]},{"index":29,"kind":"linear_attention","template":"gdn+moe_ffn","blocks":["l29.gdn","l29.moe_ffn"]},{"index":30,"kind":"linear_attention","template":"gdn+moe_ffn","blocks":["l30.gdn","l30.moe_ffn"]},{"index":31,"kind":"full_attention","template":"attn+moe_ffn","blocks":["l31.attn","l31.moe_ffn"]},{"index":32,"kind":"linear_attention","template":"gdn+moe_ffn","blocks":["l32.gdn","l32.moe_ffn"]},{"index":33,"kind":"linear_attention","template":"gdn+moe_ffn","blocks":["l33.gdn","l33.moe_ffn"]},{"index":34,"kind":"linear_attention","template":"gdn+moe_ffn","blocks":["l34.gdn","l34.moe_ffn"]},{"index":35,"kind":"full_attention","template":"attn+moe_ffn","blocks":["l35.attn","l35.moe_ffn"]},{"index":36,"kind":"linear_attention","template":"gdn+moe_ffn","blocks":["l36.gdn","l36.moe_ffn"]},{"index":37,"kind":"linear_attention","template":"gdn+moe_ffn","blocks":["l37.gdn","l37.moe_ffn"]},{"index":38,"kind":"linear_attention","template":"gdn+moe_ffn","blocks":["l38.gdn","l38.moe_ffn"]},{"index":39,"kind":"full_attention","template":"attn+moe_ffn","blocks":["l39.attn","l39.moe_ffn"]}],"nodes":[{"id":"embed.embed","order":0,"layer":null,"block":"embed","block_type":"embed","kind":null,"op":"embed","roles":[],"inserted":false,"precision":{"weight":null,"activation":null,"output":"bf16"},"steps":["gather bf16"],"group":0,"kernel":null,"family":null,"point":null,"compute":null,"numerics":"reference","where":"host","state":null},{"id":"l0.gdn.input_norm","order":1,"layer":0,"block":"l0.gdn","block_type":"gdn","kind":"linear_attention","op":"rms_norm","roles":[],"inserted":false,"precision":{"weight":null,"activation":"bf16","output":"bf16"},"steps":["compute f32"],"group":1,"kernel":"norm::rms_norm_residual","family":"rms_norm","point":null,"compute":"memory","numerics":"reference","where":"device","state":null},{"id":"l0.gdn.qkvz","order":2,"layer":0,"block":"l0.gdn","block_type":"gdn","kind":"linear_attention","op":"linear:qkvz","roles":["qkvz"],"inserted":false,"precision":{"weight":"fp8/block128x128","activation":"bf16","output":"bf16"},"steps":["act bf16","weight fp8/block128x128->bf16","mma bf16*bf16","accumulate f32","scale f32"],"group":2,"kernel":"w8a16_gemm_pipelined_m32::w8a16_gemm_pipelined_m32","family":"w8a16_gemm","point":null,"compute":"tensor_core:mma.sync.m16n8k16.bf16","numerics":"reference","where":"device","state":null},{"id":"l0.gdn.ba","order":3,"layer":0,"block":"l0.gdn","block_type":"gdn","kind":"linear_attention","op":"linear:ba","roles":["ba"],"inserted":false,"precision":{"weight":"bf16","activation":"bf16","output":"f32"},"steps":["act bf16","weight bf16->bf16","mma bf16*bf16","accumulate f32","scale none"],"group":3,"kernel":"ssm_preprocess::dense_gemv_ba_gates","family":"gdn_ba_gates_gemv","point":{"gate":"softplus","decay":"head"},"compute":"cuda_core","numerics":"reference","where":"device","state":null},{"id":"l0.gdn.gates","order":4,"layer":0,"block":"l0.gdn","block_type":"gdn","kind":"linear_attention","op":"gdn_gates","roles":[],"inserted":false,"precision":{"weight":null,"activation":"f32","output":"f32,f32"},"steps":["compute f32"],"group":3,"kernel":"ssm_preprocess::dense_gemv_ba_gates","family":"gdn_ba_gates_gemv","point":{"gate":"softplus","decay":"head"},"compute":"cuda_core","numerics":"reference","where":"device","state":null},{"id":"l0.gdn.conv","order":5,"layer":0,"block":"l0.gdn","block_type":"gdn","kind":"linear_attention","op":"conv1d_update","roles":[],"inserted":false,"precision":{"weight":null,"activation":"bf16","output":"f32"},"steps":["state f32","compute f32"],"group":4,"kernel":"causal_conv1d::causal_conv1d_update_l2norm_f32","family":"causal_conv1d_l2norm","point":null,"compute":"cuda_core","numerics":"reference","where":"device","state":"conv:update"},{"id":"l0.gdn.conv_ckpt","order":6,"layer":0,"block":"l0.gdn","block_type":"gdn","kind":"linear_attention","op":"state_snapshot","roles":[],"inserted":false,"precision":{"weight":null,"activation":"f32","output":null},"steps":["state f32"],"group":4,"kernel":"causal_conv1d::causal_conv1d_update_l2norm_f32","family":"causal_conv1d_l2norm","point":null,"compute":"cuda_core","numerics":"reference","where":"device","state":"conv:snapshot"},{"id":"l0.gdn.l2","order":7,"layer":0,"block":"l0.gdn","block_type":"gdn","kind":"linear_attention","op":"l2_norm","roles":[],"inserted":false,"precision":{"weight":null,"activation":"f32","output":"f32"},"steps":["compute f32"],"group":4,"kernel":"causal_conv1d::causal_conv1d_update_l2norm_f32","family":"causal_conv1d_l2norm","point":null,"compute":"cuda_core","numerics":"reference","where":"device","state":null},{"id":"l0.gdn.recur","order":8,"layer":0,"block":"l0.gdn","block_type":"gdn","kind":"linear_attention","op":"gdn_recurrence","roles":[],"inserted":false,"precision":{"weight":null,"activation":"f32","output":"f32"},"steps":["state f32","compute f32"],"group":5,"kernel":"gated_delta_rule::gated_delta_rule_decode_f32","family":"gdn_recurrence","point":{"decay":"head","gate_act":"silu"},"compute":"cuda_core","numerics":"reference","where":"device","state":"h:update"},{"id":"l0.gdn.out_norm","order":9,"layer":0,"block":"l0.gdn","block_type":"gdn","kind":"linear_attention","op":"gated_rms_norm","roles":[],"inserted":false,"precision":{"weight":null,"activation":"f32","output":"bf16"},"steps":["compute f32"],"group":6,"kernel":"norm::gated_rms_norm_f32_input","family":"gated_rms_norm","point":null,"compute":"memory","numerics":"reference","where":"device","state":null},{"id":"l0.gdn.out","order":10,"layer":0,"block":"l0.gdn","block_type":"gdn","kind":"linear_attention","op":"linear:gdn_out","roles":["gdn_out"],"inserted":false,"precision":{"weight":"fp8/block128x128","activation":"bf16","output":"bf16"},"steps":["act bf16","weight fp8/block128x128->bf16","mma bf16*bf16","accumulate f32","scale f32"],"group":7,"kernel":"w8a16_gemm_pipelined_m32::w8a16_gemm_pipelined_m32","family":"w8a16_gemm","point":null,"compute":"tensor_core:mma.sync.m16n8k16.bf16","numerics":"reference","where":"device","state":null},{"id":"l0.gdn.add","order":11,"layer":0,"block":"l0.gdn","block_type":"gdn","kind":"linear_attention","op":"residual_add","roles":[],"inserted":false,"precision":{"weight":null,"activation":"bf16","output":"bf16"},"steps":["compute f32"],"group":8,"kernel":"norm::residual_add_rms_norm","family":"rms_norm","point":null,"compute":"memory","numerics":"reference","where":"device","state":null},{"id":"l0.moe_ffn.post_norm","order":12,"layer":0,"block":"l0.moe_ffn","block_type":"moe_ffn","kind":"linear_attention","op":"rms_norm","roles":[],"inserted":false,"precision":{"weight":null,"activation":"f32","output":"bf16"},"steps":["compute f32"],"group":8,"kernel":"norm::residual_add_rms_norm","family":"rms_norm","point":null,"compute":"memory","numerics":"reference","where":"device","state":null},{"id":"l0.moe_ffn.router","order":13,"layer":0,"block":"l0.moe_ffn","block_type":"moe_ffn","kind":"linear_attention","op":"router","roles":[],"inserted":false,"precision":{"weight":"bf16","activation":"bf16","output":"bf16"},"steps":["act bf16","weight bf16->bf16","mma bf16*bf16","accumulate f32","scale none"],"group":9,"kernel":"gemv::dense_gemv_bf16","family":"dense_bf16","point":null,"compute":"cuda_core","numerics":"reference","where":"device","state":null},{"id":"l0.moe_ffn.top_k","order":14,"layer":0,"block":"l0.moe_ffn","block_type":"moe_ffn","kind":"linear_attention","op":"top_k","roles":[],"inserted":false,"precision":{"weight":null,"activation":"bf16","output":"f32,i32"},"steps":["score f32"],"group":10,"kernel":"moe_topk::moe_topk_softmax","family":"moe_topk","point":null,"compute":"memory","numerics":"reference","where":"device","state":null},{"id":"l0.moe_ffn.experts_gate_up","order":15,"layer":0,"block":"l0.moe_ffn","block_type":"moe_ffn","kind":"linear_attention","op":"expert_gate_up","roles":[],"inserted":false,"precision":{"weight":"fp8/block128x128","activation":"bf16","output":"bf16"},"steps":["gather bf16","act bf16","weight fp8/block128x128->bf16","mma bf16*bf16","accumulate f32","scale f32"],"group":11,"kernel":"moe_shared_expert_fused_fp8::moe_expert_gate_up_shared_fp8","family":"moe_fp8_1row","point":null,"compute":"cuda_core","numerics":"reference","where":"device","state":null},{"id":"l0.moe_ffn.shared_gate_up","order":16,"layer":0,"block":"l0.moe_ffn","block_type":"moe_ffn","kind":"linear_attention","op":"linear:shared_gate_up","roles":["shared_gate_up"],"inserted":false,"precision":{"weight":"fp8/block128x128","activation":"bf16","output":"bf16"},"steps":["act bf16","weight fp8/block128x128->bf16","mma bf16*bf16","accumulate f32","scale f32"],"group":11,"kernel":"moe_shared_expert_fused_fp8::moe_expert_gate_up_shared_fp8","family":"moe_fp8_1row","point":null,"compute":"cuda_core","numerics":"reference","where":"device","state":null},{"id":"l0.moe_ffn.experts_act","order":17,"layer":0,"block":"l0.moe_ffn","block_type":"moe_ffn","kind":"linear_attention","op":"silu_mul","roles":[],"inserted":false,"precision":{"weight":null,"activation":"bf16","output":"f32"},"steps":["compute f32"],"group":12,"kernel":"moe_shared_expert_fused_fp8::moe_expert_silu_down_shared_fp8","family":"moe_fp8_1row","point":null,"compute":"cuda_core","numerics":"reference","where":"device","state":null},{"id":"l0.moe_ffn.experts_down","order":18,"layer":0,"block":"l0.moe_ffn","block_type":"moe_ffn","kind":"linear_attention","op":"expert_down","roles":[],"inserted":false,"precision":{"weight":"fp8/block128x128","activation":"f32","output":"bf16"},"steps":["act f32","weight fp8/block128x128->f32","mma f32*f32","accumulate f32","scale f32"],"group":12,"kernel":"moe_shared_expert_fused_fp8::moe_expert_silu_down_shared_fp8","family":"moe_fp8_1row","point":null,"compute":"cuda_core","numerics":"reference","where":"device","state":null},{"id":"l0.moe_ffn.shared_act","order":19,"layer":0,"block":"l0.moe_ffn","block_type":"moe_ffn","kind":"linear_attention","op":"silu_mul","roles":[],"inserted":false,"precision":{"weight":null,"activation":"bf16","output":"f32"},"steps":["compute f32"],"group":12,"kernel":"moe_shared_expert_fused_fp8::moe_expert_silu_down_shared_fp8","family":"moe_fp8_1row","point":null,"compute":"cuda_core","numerics":"reference","where":"device","state":null},{"id":"l0.moe_ffn.shared_down","order":20,"layer":0,"block":"l0.moe_ffn","block_type":"moe_ffn","kind":"linear_attention","op":"linear:shared_down","roles":["shared_down"],"inserted":false,"precision":{"weight":"fp8/block128x128","activation":"f32","output":"bf16"},"steps":["act f32","weight fp8/block128x128->f32","mma f32*f32","accumulate f32","scale f32"],"group":12,"kernel":"moe_shared_expert_fused_fp8::moe_expert_silu_down_shared_fp8","family":"moe_fp8_1row","point":null,"compute":"cuda_core","numerics":"reference","where":"device","state":null},{"id":"l0.moe_ffn.shared_gate","order":21,"layer":0,"block":"l0.moe_ffn","block_type":"moe_ffn","kind":"linear_attention","op":"linear:shared_gate","roles":["shared_gate"],"inserted":false,"precision":{"weight":"bf16","activation":"bf16","output":"f32"},"steps":["act bf16","weight bf16->bf16","mma bf16*bf16","accumulate f32","scale none"],"group":13,"kernel":"moe_expert_gemv::moe_weighted_sum_blend","family":"moe_blend","point":{"shared_gate":"sigmoid"},"compute":"cuda_core","numerics":"reference","where":"device","state":null},{"id":"l0.moe_ffn.blend","order":22,"layer":0,"block":"l0.moe_ffn","block_type":"moe_ffn","kind":"linear_attention","op":"blend","roles":[],"inserted":false,"precision":{"weight":null,"activation":"bf16","output":"bf16"},"steps":["scatter f32","combine f32"],"group":13,"kernel":"moe_expert_gemv::moe_weighted_sum_blend","family":"moe_blend","point":{"shared_gate":"sigmoid"},"compute":"cuda_core","numerics":"reference","where":"device","state":null},{"id":"l0.moe_ffn.add","order":23,"layer":0,"block":"l0.moe_ffn","block_type":"moe_ffn","kind":"linear_attention","op":"residual_add","roles":[],"inserted":false,"precision":{"weight":null,"activation":"bf16","output":"bf16"},"steps":["compute f32"],"group":14,"kernel":"residual_add_rms_norm_exact::residual_add_rms_norm_exact","family":"rms_norm","point":null,"compute":"memory","numerics":"bit_identical(residual_add_rms_norm_exact_microtest)","where":"device","state":null},{"id":"l1.gdn.input_norm","order":24,"layer":1,"block":"l1.gdn","block_type":"gdn","kind":"linear_attention","op":"rms_norm","roles":[],"inserted":false,"precision":{"weight":null,"activation":"bf16","output":"bf16"},"steps":["compute f32"],"group":14,"kernel":"residual_add_rms_norm_exact::residual_add_rms_norm_exact","family":"rms_norm","point":null,"compute":"memory","numerics":"bit_identical(residual_add_rms_norm_exact_microtest)","where":"device","state":null},{"id":"l1.gdn.qkvz","order":25,"layer":1,"block":"l1.gdn","block_type":"gdn","kind":"linear_attention","op":"linear:qkvz","roles":["qkvz"],"inserted":false,"precision":{"weight":"fp8/block128x128","activation":"bf16","output":"bf16"},"steps":["act bf16","weight fp8/block128x128->bf16","mma bf16*bf16","accumulate f32","scale f32"],"group":15,"kernel":"w8a16_gemm_pipelined_m32::w8a16_gemm_pipelined_m32","family":"w8a16_gemm","point":null,"compute":"tensor_core:mma.sync.m16n8k16.bf16","numerics":"reference","where":"device","state":null},{"id":"l1.gdn.ba","order":26,"layer":1,"block":"l1.gdn","block_type":"gdn","kind":"linear_attention","op":"linear:ba","roles":["ba"],"inserted":false,"precision":{"weight":"bf16","activation":"bf16","output":"f32"},"steps":["act bf16","weight bf16->bf16","mma bf16*bf16","accumulate f32","scale none"],"group":16,"kernel":"ssm_preprocess::dense_gemv_ba_gates","family":"gdn_ba_gates_gemv","point":{"gate":"softplus","decay":"head"},"compute":"cuda_core","numerics":"reference","where":"device","state":null},{"id":"l1.gdn.gates","order":27,"layer":1,"block":"l1.gdn","block_type":"gdn","kind":"linear_attention","op":"gdn_gates","roles":[],"inserted":false,"precision":{"weight":null,"activation":"f32","output":"f32,f32"},"steps":["compute f32"],"group":16,"kernel":"ssm_preprocess::dense_gemv_ba_gates","family":"gdn_ba_gates_gemv","point":{"gate":"softplus","decay":"head"},"compute":"cuda_core","numerics":"reference","where":"device","state":null},{"id":"l1.gdn.conv","order":28,"layer":1,"block":"l1.gdn","block_type":"gdn","kind":"linear_attention","op":"conv1d_update","roles":[],"inserted":false,"precision":{"weight":null,"activation":"bf16","output":"f32"},"steps":["state f32","compute f32"],"group":17,"kernel":"causal_conv1d::causal_conv1d_update_l2norm_f32","family":"causal_conv1d_l2norm","point":null,"compute":"cuda_core","numerics":"reference","where":"device","state":"conv:update"},{"id":"l1.gdn.conv_ckpt","order":29,"layer":1,"block":"l1.gdn","block_type":"gdn","kind":"linear_attention","op":"state_snapshot","roles":[],"inserted":false,"precision":{"weight":null,"activation":"f32","output":null},"steps":["state f32"],"group":17,"kernel":"causal_conv1d::causal_conv1d_update_l2norm_f32","family":"causal_conv1d_l2norm","point":null,"compute":"cuda_core","numerics":"reference","where":"device","state":"conv:snapshot"},{"id":"l1.gdn.l2","order":30,"layer":1,"block":"l1.gdn","block_type":"gdn","kind":"linear_attention","op":"l2_norm","roles":[],"inserted":false,"precision":{"weight":null,"activation":"f32","output":"f32"},"steps":["compute f32"],"group":17,"kernel":"causal_conv1d::causal_conv1d_update_l2norm_f32","family":"causal_conv1d_l2norm","point":null,"compute":"cuda_core","numerics":"reference","where":"device","state":null},{"id":"l1.gdn.recur","order":31,"layer":1,"block":"l1.gdn","block_type":"gdn","kind":"linear_attention","op":"gdn_recurrence","roles":[],"inserted":false,"precision":{"weight":null,"activation":"f32","output":"f32"},"steps":["state f32","compute f32"],"group":18,"kernel":"gated_delta_rule::gated_delta_rule_decode_f32","family":"gdn_recurrence","point":{"decay":"head","gate_act":"silu"},"compute":"cuda_core","numerics":"reference","where":"device","state":"h:update"},{"id":"l1.gdn.out_norm","order":32,"layer":1,"block":"l1.gdn","block_type":"gdn","kind":"linear_attention","op":"gated_rms_norm","roles":[],"inserted":false,"precision":{"weight":null,"activation":"f32","output":"bf16"},"steps":["compute f32"],"group":19,"kernel":"norm::gated_rms_norm_f32_input","family":"gated_rms_norm","point":null,"compute":"memory","numerics":"reference","where":"device","state":null},{"id":"l1.gdn.out","order":33,"layer":1,"block":"l1.gdn","block_type":"gdn","kind":"linear_attention","op":"linear:gdn_out","roles":["gdn_out"],"inserted":false,"precision":{"weight":"fp8/block128x128","activation":"bf16","output":"bf16"},"steps":["act bf16","weight fp8/block128x128->bf16","mma bf16*bf16","accumulate f32","scale f32"],"group":20,"kernel":"w8a16_gemm_pipelined_m32::w8a16_gemm_pipelined_m32","family":"w8a16_gemm","point":null,"compute":"tensor_core:mma.sync.m16n8k16.bf16","numerics":"reference","where":"device","state":null},{"id":"l1.gdn.add","order":34,"layer":1,"block":"l1.gdn","block_type":"gdn","kind":"linear_attention","op":"residual_add","roles":[],"inserted":false,"precision":{"weight":null,"activation":"bf16","output":"bf16"},"steps":["compute f32"],"group":21,"kernel":"norm::residual_add_rms_norm","family":"rms_norm","point":null,"compute":"memory","numerics":"reference","where":"device","state":null},{"id":"l1.moe_ffn.post_norm","order":35,"layer":1,"block":"l1.moe_ffn","block_type":"moe_ffn","kind":"linear_attention","op":"rms_norm","roles":[],"inserted":false,"precision":{"weight":null,"activation":"f32","output":"bf16"},"steps":["compute f32"],"group":21,"kernel":"norm::residual_add_rms_norm","family":"rms_norm","point":null,"compute":"memory","numerics":"reference","where":"device","state":null},{"id":"l1.moe_ffn.router","order":36,"layer":1,"block":"l1.moe_ffn","block_type":"moe_ffn","kind":"linear_attention","op":"router","roles":[],"inserted":false,"precision":{"weight":"bf16","activation":"bf16","output":"bf16"},"steps":["act bf16","weight bf16->bf16","mma bf16*bf16","accumulate f32","scale none"],"group":22,"kernel":"gemv::dense_gemv_bf16","family":"dense_bf16","point":null,"compute":"cuda_core","numerics":"reference","where":"device","state":null},{"id":"l1.moe_ffn.top_k","order":37,"layer":1,"block":"l1.moe_ffn","block_type":"moe_ffn","kind":"linear_attention","op":"top_k","roles":[],"inserted":false,"precision":{"weight":null,"activation":"bf16","output":"f32,i32"},"steps":["score f32"],"group":23,"kernel":"moe_topk::moe_topk_softmax","family":"moe_topk","point":null,"compute":"memory","numerics":"reference","where":"device","state":null},{"id":"l1.moe_ffn.experts_gate_up","order":38,"layer":1,"block":"l1.moe_ffn","block_type":"moe_ffn","kind":"linear_attention","op":"expert_gate_up","roles":[],"inserted":false,"precision":{"weight":"fp8/block128x128","activation":"bf16","output":"bf16"},"steps":["gather bf16","act bf16","weight fp8/block128x128->bf16","mma bf16*bf16","accumulate f32","scale f32"],"group":24,"kernel":"moe_shared_expert_fused_fp8::moe_expert_gate_up_shared_fp8","family":"moe_fp8_1row","point":null,"compute":"cuda_core","numerics":"reference","where":"device","state":null},{"id":"l1.moe_ffn.shared_gate_up","order":39,"layer":1,"block":"l1.moe_ffn","block_type":"moe_ffn","kind":"linear_attention","op":"linear:shared_gate_up","roles":["shared_gate_up"],"inserted":false,"precision":{"weight":"fp8/block128x128","activation":"bf16","output":"bf16"},"steps":["act bf16","weight fp8/block128x128->bf16","mma bf16*bf16","accumulate f32","scale f32"],"group":24,"kernel":"moe_shared_expert_fused_fp8::moe_expert_gate_up_shared_fp8","family":"moe_fp8_1row","point":null,"compute":"cuda_core","numerics":"reference","where":"device","state":null},{"id":"l1.moe_ffn.experts_act","order":40,"layer":1,"block":"l1.moe_ffn","block_type":"moe_ffn","kind":"linear_attention","op":"silu_mul","roles":[],"inserted":false,"precision":{"weight":null,"activation":"bf16","output":"f32"},"steps":["compute f32"],"group":25,"kernel":"moe_shared_expert_fused_fp8::moe_expert_silu_down_shared_fp8","family":"moe_fp8_1row","point":null,"compute":"cuda_core","numerics":"reference","where":"device","state":null},{"id":"l1.moe_ffn.experts_down","order":41,"layer":1,"block":"l1.moe_ffn","block_type":"moe_ffn","kind":"linear_attention","op":"expert_down","roles":[],"inserted":false,"precision":{"weight":"fp8/block128x128","activation":"f32","output":"bf16"},"steps":["act f32","weight fp8/block128x128->f32","mma f32*f32","accumulate f32","scale f32"],"group":25,"kernel":"moe_shared_expert_fused_fp8::moe_expert_silu_down_shared_fp8","family":"moe_fp8_1row","point":null,"compute":"cuda_core","numerics":"reference","where":"device","state":null},{"id":"l1.moe_ffn.shared_act","order":42,"layer":1,"block":"l1.moe_ffn","block_type":"moe_ffn","kind":"linear_attention","op":"silu_mul","roles":[],"inserted":false,"precision":{"weight":null,"activation":"bf16","output":"f32"},"steps":["compute f32"],"group":25,"kernel":"moe_shared_expert_fused_fp8::moe_expert_silu_down_shared_fp8","family":"moe_fp8_1row","point":null,"compute":"cuda_core","numerics":"reference","where":"device","state":null},{"id":"l1.moe_ffn.shared_down","order":43,"layer":1,"block":"l1.moe_ffn","block_type":"moe_ffn","kind":"linear_attention","op":"linear:shared_down","roles":["shared_down"],"inserted":false,"precision":{"weight":"fp8/block128x128","activation":"f32","output":"bf16"},"steps":["act f32","weight fp8/block128x128->f32","mma f32*f32","accumulate f32","scale f32"],"group":25,"kernel":"moe_shared_expert_fused_fp8::moe_expert_silu_down_shared_fp8","family":"moe_fp8_1row","point":null,"compute":"cuda_core","numerics":"reference","where":"device","state":null},{"id":"l1.moe_ffn.shared_gate","order":44,"layer":1,"block":"l1.moe_ffn","block_type":"moe_ffn","kind":"linear_attention","op":"linear:shared_gate","roles":["shared_gate"],"inserted":false,"precision":{"weight":"bf16","activation":"bf16","output":"f32"},"steps":["act bf16","weight bf16->bf16","mma bf16*bf16","accumulate f32","scale none"],"group":26,"kernel":"moe_expert_gemv::moe_weighted_sum_blend","family":"moe_blend","point":{"shared_gate":"sigmoid"},"compute":"cuda_core","numerics":"reference","where":"device","state":null},{"id":"l1.moe_ffn.blend","order":45,"layer":1,"block":"l1.moe_ffn","block_type":"moe_ffn","kind":"linear_attention","op":"blend","roles":[],"inserted":false,"precision":{"weight":null,"activation":"bf16","output":"bf16"},"steps":["scatter f32","combine f32"],"group":26,"kernel":"moe_expert_gemv::moe_weighted_sum_blend","family":"moe_blend","point":{"shared_gate":"sigmoid"},"compute":"cuda_core","numerics":"reference","where":"device","state":null},{"id":"l1.moe_ffn.add","order":46,"layer":1,"block":"l1.moe_ffn","block_type":"moe_ffn","kind":"linear_attention","op":"residual_add","roles":[],"inserted":false,"precision":{"weight":null,"activation":"bf16","output":"bf16"},"steps":["compute f32"],"group":27,"kernel":"residual_add_rms_norm_exact::residual_add_rms_norm_exact","family":"rms_norm","point":null,"compute":"memory","numerics":"bit_identical(residual_add_rms_norm_exact_microtest)","where":"device","state":null},{"id":"l2.gdn.input_norm","order":47,"layer":2,"block":"l2.gdn","block_type":"gdn","kind":"linear_attention","op":"rms_norm","roles":[],"inserted":false,"precision":{"weight":null,"activation":"bf16","output":"bf16"},"steps":["compute f32"],"group":27,"kernel":"residual_add_rms_norm_exact::residual_add_rms_norm_exact","family":"rms_norm","point":null,"compute":"memory","numerics":"bit_identical(residual_add_rms_norm_exact_microtest)","where":"device","state":null},{"id":"l2.gdn.qkvz","order":48,"layer":2,"block":"l2.gdn","block_type":"gdn","kind":"linear_attention","op":"linear:qkvz","roles":["qkvz"],"inserted":false,"precision":{"weight":"fp8/block128x128","activation":"bf16","output":"bf16"},"steps":["act bf16","weight fp8/block128x128->bf16","mma bf16*bf16","accumulate f32","scale f32"],"group":28,"kernel":"w8a16_gemm_pipelined_m32::w8a16_gemm_pipelined_m32","family":"w8a16_gemm","point":null,"compute":"tensor_core:mma.sync.m16n8k16.bf16","numerics":"reference","where":"device","state":null},{"id":"l2.gdn.ba","order":49,"layer":2,"block":"l2.gdn","block_type":"gdn","kind":"linear_attention","op":"linear:ba","roles":["ba"],"inserted":false,"precision":{"weight":"bf16","activation":"bf16","output":"f32"},"steps":["act bf16","weight bf16->bf16","mma bf16*bf16","accumulate f32","scale none"],"group":29,"kernel":"ssm_preprocess::dense_gemv_ba_gates","family":"gdn_ba_gates_gemv","point":{"gate":"softplus","decay":"head"},"compute":"cuda_core","numerics":"reference","where":"device","state":null},{"id":"l2.gdn.gates","order":50,"layer":2,"block":"l2.gdn","block_type":"gdn","kind":"linear_attention","op":"gdn_gates","roles":[],"inserted":false,"precision":{"weight":null,"activation":"f32","output":"f32,f32"},"steps":["compute f32"],"group":29,"kernel":"ssm_preprocess::dense_gemv_ba_gates","family":"gdn_ba_gates_gemv","point":{"gate":"softplus","decay":"head"},"compute":"cuda_core","numerics":"reference","where":"device","state":null},{"id":"l2.gdn.conv","order":51,"layer":2,"block":"l2.gdn","block_type":"gdn","kind":"linear_attention","op":"conv1d_update","roles":[],"inserted":false,"precision":{"weight":null,"activation":"bf16","output":"f32"},"steps":["state f32","compute f32"],"group":30,"kernel":"causal_conv1d::causal_conv1d_update_l2norm_f32","family":"causal_conv1d_l2norm","point":null,"compute":"cuda_core","numerics":"reference","where":"device","state":"conv:update"},{"id":"l2.gdn.conv_ckpt","order":52,"layer":2,"block":"l2.gdn","block_type":"gdn","kind":"linear_attention","op":"state_snapshot","roles":[],"inserted":false,"precision":{"weight":null,"activation":"f32","output":null},"steps":["state f32"],"group":30,"kernel":"causal_conv1d::causal_conv1d_update_l2norm_f32","family":"causal_conv1d_l2norm","point":null,"compute":"cuda_core","numerics":"reference","where":"device","state":"conv:snapshot"},{"id":"l2.gdn.l2","order":53,"layer":2,"block":"l2.gdn","block_type":"gdn","kind":"linear_attention","op":"l2_norm","roles":[],"inserted":false,"precision":{"weight":null,"activation":"f32","output":"f32"},"steps":["compute f32"],"group":30,"kernel":"causal_conv1d::causal_conv1d_update_l2norm_f32","family":"causal_conv1d_l2norm","point":null,"compute":"cuda_core","numerics":"reference","where":"device","state":null},{"id":"l2.gdn.recur","order":54,"layer":2,"block":"l2.gdn","block_type":"gdn","kind":"linear_attention","op":"gdn_recurrence","roles":[],"inserted":false,"precision":{"weight":null,"activation":"f32","output":"f32"},"steps":["state f32","compute f32"],"group":31,"kernel":"gated_delta_rule::gated_delta_rule_decode_f32","family":"gdn_recurrence","point":{"decay":"head","gate_act":"silu"},"compute":"cuda_core","numerics":"reference","where":"device","state":"h:update"},{"id":"l2.gdn.out_norm","order":55,"layer":2,"block":"l2.gdn","block_type":"gdn","kind":"linear_attention","op":"gated_rms_norm","roles":[],"inserted":false,"precision":{"weight":null,"activation":"f32","output":"bf16"},"steps":["compute f32"],"group":32,"kernel":"norm::gated_rms_norm_f32_input","family":"gated_rms_norm","point":null,"compute":"memory","numerics":"reference","where":"device","state":null},{"id":"l2.gdn.out","order":56,"layer":2,"block":"l2.gdn","block_type":"gdn","kind":"linear_attention","op":"linear:gdn_out","roles":["gdn_out"],"inserted":false,"precision":{"weight":"fp8/block128x128","activation":"bf16","output":"bf16"},"steps":["act bf16","weight fp8/block128x128->bf16","mma bf16*bf16","accumulate f32","scale f32"],"group":33,"kernel":"w8a16_gemm_pipelined_m32::w8a16_gemm_pipelined_m32","family":"w8a16_gemm","point":null,"compute":"tensor_core:mma.sync.m16n8k16.bf16","numerics":"reference","where":"device","state":null},{"id":"l2.gdn.add","order":57,"layer":2,"block":"l2.gdn","block_type":"gdn","kind":"linear_attention","op":"residual_add","roles":[],"inserted":false,"precision":{"weight":null,"activation":"bf16","output":"bf16"},"steps":["compute f32"],"group":34,"kernel":"norm::residual_add_rms_norm","family":"rms_norm","point":null,"compute":"memory","numerics":"reference","where":"device","state":null},{"id":"l2.moe_ffn.post_norm","order":58,"layer":2,"block":"l2.moe_ffn","block_type":"moe_ffn","kind":"linear_attention","op":"rms_norm","roles":[],"inserted":false,"precision":{"weight":null,"activation":"f32","output":"bf16"},"steps":["compute f32"],"group":34,"kernel":"norm::residual_add_rms_norm","family":"rms_norm","point":null,"compute":"memory","numerics":"reference","where":"device","state":null},{"id":"l2.moe_ffn.router","order":59,"layer":2,"block":"l2.moe_ffn","block_type":"moe_ffn","kind":"linear_attention","op":"router","roles":[],"inserted":false,"precision":{"weight":"bf16","activation":"bf16","output":"bf16"},"steps":["act bf16","weight bf16->bf16","mma bf16*bf16","accumulate f32","scale none"],"group":35,"kernel":"gemv::dense_gemv_bf16","family":"dense_bf16","point":null,"compute":"cuda_core","numerics":"reference","where":"device","state":null},{"id":"l2.moe_ffn.top_k","order":60,"layer":2,"block":"l2.moe_ffn","block_type":"moe_ffn","kind":"linear_attention","op":"top_k","roles":[],"inserted":false,"precision":{"weight":null,"activation":"bf16","output":"f32,i32"},"steps":["score f32"],"group":36,"kernel":"moe_topk::moe_topk_softmax","family":"moe_topk","point":null,"compute":"memory","numerics":"reference","where":"device","state":null},{"id":"l2.moe_ffn.experts_gate_up","order":61,"layer":2,"block":"l2.moe_ffn","block_type":"moe_ffn","kind":"linear_attention","op":"expert_gate_up","roles":[],"inserted":false,"precision":{"weight":"fp8/block128x128","activation":"bf16","output":"bf16"},"steps":["gather bf16","act bf16","weight fp8/block128x128->bf16","mma bf16*bf16","accumulate f32","scale f32"],"group":37,"kernel":"moe_shared_expert_fused_fp8::moe_expert_gate_up_shared_fp8","family":"moe_fp8_1row","point":null,"compute":"cuda_core","numerics":"reference","where":"device","state":null},{"id":"l2.moe_ffn.shared_gate_up","order":62,"layer":2,"block":"l2.moe_ffn","block_type":"moe_ffn","kind":"linear_attention","op":"linear:shared_gate_up","roles":["shared_gate_up"],"inserted":false,"precision":{"weight":"fp8/block128x128","activation":"bf16","output":"bf16"},"steps":["act bf16","weight fp8/block128x128->bf16","mma bf16*bf16","accumulate f32","scale f32"],"group":37,"kernel":"moe_shared_expert_fused_fp8::moe_expert_gate_up_shared_fp8","family":"moe_fp8_1row","point":null,"compute":"cuda_core","numerics":"reference","where":"device","state":null},{"id":"l2.moe_ffn.experts_act","order":63,"layer":2,"block":"l2.moe_ffn","block_type":"moe_ffn","kind":"linear_attention","op":"silu_mul","roles":[],"inserted":false,"precision":{"weight":null,"activation":"bf16","output":"f32"},"steps":["compute f32"],"group":38,"kernel":"moe_shared_expert_fused_fp8::moe_expert_silu_down_shared_fp8","family":"moe_fp8_1row","point":null,"compute":"cuda_core","numerics":"reference","where":"device","state":null},{"id":"l2.moe_ffn.experts_down","order":64,"layer":2,"block":"l2.moe_ffn","block_type":"moe_ffn","kind":"linear_attention","op":"expert_down","roles":[],"inserted":false,"precision":{"weight":"fp8/block128x128","activation":"f32","output":"bf16"},"steps":["act f32","weight fp8/block128x128->f32","mma f32*f32","accumulate f32","scale f32"],"group":38,"kernel":"moe_shared_expert_fused_fp8::moe_expert_silu_down_shared_fp8","family":"moe_fp8_1row","point":null,"compute":"cuda_core","numerics":"reference","where":"device","state":null},{"id":"l2.moe_ffn.shared_act","order":65,"layer":2,"block":"l2.moe_ffn","block_type":"moe_ffn","kind":"linear_attention","op":"silu_mul","roles":[],"inserted":false,"precision":{"weight":null,"activation":"bf16","output":"f32"},"steps":["compute f32"],"group":38,"kernel":"moe_shared_expert_fused_fp8::moe_expert_silu_down_shared_fp8","family":"moe_fp8_1row","point":null,"compute":"cuda_core","numerics":"reference","where":"device","state":null},{"id":"l2.moe_ffn.shared_down","order":66,"layer":2,"block":"l2.moe_ffn","block_type":"moe_ffn","kind":"linear_attention","op":"linear:shared_down","roles":["shared_down"],"inserted":false,"precision":{"weight":"fp8/block128x128","activation":"f32","output":"bf16"},"steps":["act f32","weight fp8/block128x128->f32","mma f32*f32","accumulate f32","scale f32"],"group":38,"kernel":"moe_shared_expert_fused_fp8::moe_expert_silu_down_shared_fp8","family":"moe_fp8_1row","point":null,"compute":"cuda_core","numerics":"reference","where":"device","state":null},{"id":"l2.moe_ffn.shared_gate","order":67,"layer":2,"block":"l2.moe_ffn","block_type":"moe_ffn","kind":"linear_attention","op":"linear:shared_gate","roles":["shared_gate"],"inserted":false,"precision":{"weight":"bf16","activation":"bf16","output":"f32"},"steps":["act bf16","weight bf16->bf16","mma bf16*bf16","accumulate f32","scale none"],"group":39,"kernel":"moe_expert_gemv::moe_weighted_sum_blend","family":"moe_blend","point":{"shared_gate":"sigmoid"},"compute":"cuda_core","numerics":"reference","where":"device","state":null},{"id":"l2.moe_ffn.blend","order":68,"layer":2,"block":"l2.moe_ffn","block_type":"moe_ffn","kind":"linear_attention","op":"blend","roles":[],"inserted":false,"precision":{"weight":null,"activation":"bf16","output":"bf16"},"steps":["scatter f32","combine f32"],"group":39,"kernel":"moe_expert_gemv::moe_weighted_sum_blend","family":"moe_blend","point":{"shared_gate":"sigmoid"},"compute":"cuda_core","numerics":"reference","where":"device","state":null},{"id":"l2.moe_ffn.add","order":69,"layer":2,"block":"l2.moe_ffn","block_type":"moe_ffn","kind":"linear_attention","op":"residual_add","roles":[],"inserted":false,"precision":{"weight":null,"activation":"bf16","output":"bf16"},"steps":["compute f32"],"group":40,"kernel":"residual_add_rms_norm_exact::residual_add_rms_norm_exact","family":"rms_norm","point":null,"compute":"memory","numerics":"bit_identical(residual_add_rms_norm_exact_microtest)","where":"device","state":null},{"id":"l3.attn.input_norm","order":70,"layer":3,"block":"l3.attn","block_type":"attn","kind":"full_attention","op":"rms_norm","roles":[],"inserted":false,"precision":{"weight":null,"activation":"bf16","output":"bf16"},"steps":["compute f32"],"group":40,"kernel":"residual_add_rms_norm_exact::residual_add_rms_norm_exact","family":"rms_norm","point":null,"compute":"memory","numerics":"bit_identical(residual_add_rms_norm_exact_microtest)","where":"device","state":null},{"id":"l3.attn.q","order":71,"layer":3,"block":"l3.attn","block_type":"attn","kind":"full_attention","op":"linear:q","roles":["q"],"inserted":false,"precision":{"weight":"fp8/block128x128","activation":"bf16","output":"bf16"},"steps":["act bf16","weight fp8/block128x128->bf16","mma bf16*bf16","accumulate f32","scale f32"],"group":41,"kernel":"w8a16_gemm_pipelined_m32::w8a16_gemm_pipelined_m32","family":"w8a16_gemm","point":null,"compute":"tensor_core:mma.sync.m16n8k16.bf16","numerics":"reference","where":"device","state":null},{"id":"l3.attn.q_split","order":72,"layer":3,"block":"l3.attn","block_type":"attn","kind":"full_attention","op":"split","roles":[],"inserted":false,"precision":{"weight":null,"activation":"bf16","output":"bf16,bf16"},"steps":["move bf16"],"group":42,"kernel":"ssm_preprocess::deinterleave_qg","family":"deinterleave_qg","point":{"layout":"interleaved"},"compute":"memory","numerics":"reference","where":"device","state":null},{"id":"l3.attn.k","order":73,"layer":3,"block":"l3.attn","block_type":"attn","kind":"full_attention","op":"linear:k","roles":["k"],"inserted":false,"precision":{"weight":"fp8/block128x128","activation":"bf16","output":"bf16"},"steps":["act bf16","weight fp8/block128x128->bf16","mma bf16*bf16","accumulate f32","scale f32"],"group":43,"kernel":"w8a16_gemm_pipelined_m32::w8a16_gemm_pipelined_m32","family":"w8a16_gemm","point":null,"compute":"tensor_core:mma.sync.m16n8k16.bf16","numerics":"reference","where":"device","state":null},{"id":"l3.attn.v","order":74,"layer":3,"block":"l3.attn","block_type":"attn","kind":"full_attention","op":"linear:v","roles":["v"],"inserted":false,"precision":{"weight":"fp8/block128x128","activation":"bf16","output":"bf16"},"steps":["act bf16","weight fp8/block128x128->bf16","mma bf16*bf16","accumulate f32","scale f32"],"group":44,"kernel":"w8a16_gemm_pipelined_m32::w8a16_gemm_pipelined_m32","family":"w8a16_gemm","point":null,"compute":"tensor_core:mma.sync.m16n8k16.bf16","numerics":"reference","where":"device","state":null},{"id":"l3.attn.q_norm","order":75,"layer":3,"block":"l3.attn","block_type":"attn","kind":"full_attention","op":"qk_norm","roles":[],"inserted":false,"precision":{"weight":null,"activation":"bf16","output":"bf16"},"steps":["compute f32"],"group":45,"kernel":"norm::rms_norm","family":"rms_norm","point":null,"compute":"memory","numerics":"reference","where":"device","state":null},{"id":"l3.attn.k_norm","order":76,"layer":3,"block":"l3.attn","block_type":"attn","kind":"full_attention","op":"qk_norm","roles":[],"inserted":false,"precision":{"weight":null,"activation":"bf16","output":"bf16"},"steps":["compute f32"],"group":46,"kernel":"norm::rms_norm","family":"rms_norm","point":null,"compute":"memory","numerics":"reference","where":"device","state":null},{"id":"l3.attn.rope","order":77,"layer":3,"block":"l3.attn","block_type":"attn","kind":"full_attention","op":"rope","roles":[],"inserted":false,"precision":{"weight":null,"activation":"bf16","output":"bf16,bf16"},"steps":["compute f32"],"group":47,"kernel":"rope_mrope_interleaved::rope_forward_mrope_interleaved","family":"rope","point":null,"compute":"memory","numerics":"reference","where":"device","state":null},{"id":"l3.attn.kv_write","order":78,"layer":3,"block":"l3.attn","block_type":"attn","kind":"full_attention","op":"kv_write","roles":[],"inserted":false,"precision":{"weight":null,"activation":"bf16","output":null},"steps":["cache bf16"],"group":48,"kernel":"reshape_and_cache::reshape_and_cache_flash","family":"kv_write","point":null,"compute":"memory","numerics":"reference","where":"device","state":"k:write,v:write"},{"id":"l3.attn.attend","order":79,"layer":3,"block":"l3.attn","block_type":"attn","kind":"full_attention","op":"paged_attention","roles":[],"inserted":false,"precision":{"weight":null,"activation":"bf16","output":"bf16"},"steps":["cache bf16","scores f32","softmax f32","accumulate f32"],"group":49,"kernel":"paged_decode::paged_decode_attn","family":"paged_decode_attn","point":null,"compute":"cuda_core","numerics":"reference","where":"device","state":"k:read,v:read"},{"id":"l3.attn.gate_mul","order":80,"layer":3,"block":"l3.attn","block_type":"attn","kind":"full_attention","op":"sigmoid_gate_mul","roles":[],"inserted":false,"precision":{"weight":null,"activation":"bf16","output":"bf16"},"steps":["compute f32"],"group":50,"kernel":"residual_add::sigmoid_gate_mul","family":"sigmoid_gate_mul","point":null,"compute":"memory","numerics":"reference","where":"device","state":null},{"id":"l3.attn.o","order":81,"layer":3,"block":"l3.attn","block_type":"attn","kind":"full_attention","op":"linear:o","roles":["o"],"inserted":false,"precision":{"weight":"fp8/block128x128","activation":"bf16","output":"bf16"},"steps":["act bf16","weight fp8/block128x128->bf16","mma bf16*bf16","accumulate f32","scale f32"],"group":51,"kernel":"w8a16_gemm_pipelined_m32::w8a16_gemm_pipelined_m32","family":"w8a16_gemm","point":null,"compute":"tensor_core:mma.sync.m16n8k16.bf16","numerics":"reference","where":"device","state":null},{"id":"l3.attn.add","order":82,"layer":3,"block":"l3.attn","block_type":"attn","kind":"full_attention","op":"residual_add","roles":[],"inserted":false,"precision":{"weight":null,"activation":"bf16","output":"bf16"},"steps":["compute f32"],"group":52,"kernel":"norm::residual_add_rms_norm","family":"rms_norm","point":null,"compute":"memory","numerics":"reference","where":"device","state":null},{"id":"l3.moe_ffn.post_norm","order":83,"layer":3,"block":"l3.moe_ffn","block_type":"moe_ffn","kind":"full_attention","op":"rms_norm","roles":[],"inserted":false,"precision":{"weight":null,"activation":"f32","output":"bf16"},"steps":["compute f32"],"group":52,"kernel":"norm::residual_add_rms_norm","family":"rms_norm","point":null,"compute":"memory","numerics":"reference","where":"device","state":null},{"id":"l3.moe_ffn.router","order":84,"layer":3,"block":"l3.moe_ffn","block_type":"moe_ffn","kind":"full_attention","op":"router","roles":[],"inserted":false,"precision":{"weight":"bf16","activation":"bf16","output":"bf16"},"steps":["act bf16","weight bf16->bf16","mma bf16*bf16","accumulate f32","scale none"],"group":53,"kernel":"gemv::dense_gemv_bf16","family":"dense_bf16","point":null,"compute":"cuda_core","numerics":"reference","where":"device","state":null},{"id":"l3.moe_ffn.top_k","order":85,"layer":3,"block":"l3.moe_ffn","block_type":"moe_ffn","kind":"full_attention","op":"top_k","roles":[],"inserted":false,"precision":{"weight":null,"activation":"bf16","output":"f32,i32"},"steps":["score f32"],"group":54,"kernel":"moe_topk::moe_topk_softmax","family":"moe_topk","point":null,"compute":"memory","numerics":"reference","where":"device","state":null},{"id":"l3.moe_ffn.experts_gate_up","order":86,"layer":3,"block":"l3.moe_ffn","block_type":"moe_ffn","kind":"full_attention","op":"expert_gate_up","roles":[],"inserted":false,"precision":{"weight":"fp8/block128x128","activation":"bf16","output":"bf16"},"steps":["gather bf16","act bf16","weight fp8/block128x128->bf16","mma bf16*bf16","accumulate f32","scale f32"],"group":55,"kernel":"moe_shared_expert_fused_fp8::moe_expert_gate_up_shared_fp8","family":"moe_fp8_1row","point":null,"compute":"cuda_core","numerics":"reference","where":"device","state":null},{"id":"l3.moe_ffn.shared_gate_up","order":87,"layer":3,"block":"l3.moe_ffn","block_type":"moe_ffn","kind":"full_attention","op":"linear:shared_gate_up","roles":["shared_gate_up"],"inserted":false,"precision":{"weight":"fp8/block128x128","activation":"bf16","output":"bf16"},"steps":["act bf16","weight fp8/block128x128->bf16","mma bf16*bf16","accumulate f32","scale f32"],"group":55,"kernel":"moe_shared_expert_fused_fp8::moe_expert_gate_up_shared_fp8","family":"moe_fp8_1row","point":null,"compute":"cuda_core","numerics":"reference","where":"device","state":null},{"id":"l3.moe_ffn.experts_act","order":88,"layer":3,"block":"l3.moe_ffn","block_type":"moe_ffn","kind":"full_attention","op":"silu_mul","roles":[],"inserted":false,"precision":{"weight":null,"activation":"bf16","output":"f32"},"steps":["compute f32"],"group":56,"kernel":"moe_shared_expert_fused_fp8::moe_expert_silu_down_shared_fp8","family":"moe_fp8_1row","point":null,"compute":"cuda_core","numerics":"reference","where":"device","state":null},{"id":"l3.moe_ffn.experts_down","order":89,"layer":3,"block":"l3.moe_ffn","block_type":"moe_ffn","kind":"full_attention","op":"expert_down","roles":[],"inserted":false,"precision":{"weight":"fp8/block128x128","activation":"f32","output":"bf16"},"steps":["act f32","weight fp8/block128x128->f32","mma f32*f32","accumulate f32","scale f32"],"group":56,"kernel":"moe_shared_expert_fused_fp8::moe_expert_silu_down_shared_fp8","family":"moe_fp8_1row","point":null,"compute":"cuda_core","numerics":"reference","where":"device","state":null},{"id":"l3.moe_ffn.shared_act","order":90,"layer":3,"block":"l3.moe_ffn","block_type":"moe_ffn","kind":"full_attention","op":"silu_mul","roles":[],"inserted":false,"precision":{"weight":null,"activation":"bf16","output":"f32"},"steps":["compute f32"],"group":56,"kernel":"moe_shared_expert_fused_fp8::moe_expert_silu_down_shared_fp8","family":"moe_fp8_1row","point":null,"compute":"cuda_core","numerics":"reference","where":"device","state":null},{"id":"l3.moe_ffn.shared_down","order":91,"layer":3,"block":"l3.moe_ffn","block_type":"moe_ffn","kind":"full_attention","op":"linear:shared_down","roles":["shared_down"],"inserted":false,"precision":{"weight":"fp8/block128x128","activation":"f32","output":"bf16"},"steps":["act f32","weight fp8/block128x128->f32","mma f32*f32","accumulate f32","scale f32"],"group":56,"kernel":"moe_shared_expert_fused_fp8::moe_expert_silu_down_shared_fp8","family":"moe_fp8_1row","point":null,"compute":"cuda_core","numerics":"reference","where":"device","state":null},{"id":"l3.moe_ffn.shared_gate","order":92,"layer":3,"block":"l3.moe_ffn","block_type":"moe_ffn","kind":"full_attention","op":"linear:shared_gate","roles":["shared_gate"],"inserted":false,"precision":{"weight":"bf16","activation":"bf16","output":"f32"},"steps":["act bf16","weight bf16->bf16","mma bf16*bf16","accumulate f32","scale none"],"group":57,"kernel":"moe_expert_gemv::moe_weighted_sum_blend","family":"moe_blend","point":{"shared_gate":"sigmoid"},"compute":"cuda_core","numerics":"reference","where":"device","state":null},{"id":"l3.moe_ffn.blend","order":93,"layer":3,"block":"l3.moe_ffn","block_type":"moe_ffn","kind":"full_attention","op":"blend","roles":[],"inserted":false,"precision":{"weight":null,"activation":"bf16","output":"bf16"},"steps":["scatter f32","combine f32"],"group":57,"kernel":"moe_expert_gemv::moe_weighted_sum_blend","family":"moe_blend","point":{"shared_gate":"sigmoid"},"compute":"cuda_core","numerics":"reference","where":"device","state":null},{"id":"l3.moe_ffn.add","order":94,"layer":3,"block":"l3.moe_ffn","block_type":"moe_ffn","kind":"full_attention","op":"residual_add","roles":[],"inserted":false,"precision":{"weight":null,"activation":"bf16","output":"bf16"},"steps":["compute f32"],"group":58,"kernel":"residual_add_rms_norm_exact::residual_add_rms_norm_exact","family":"rms_norm","point":null,"compute":"memory","numerics":"bit_identical(residual_add_rms_norm_exact_microtest)","where":"device","state":null},{"id":"l4.gdn.input_norm","order":95,"layer":4,"block":"l4.gdn","block_type":"gdn","kind":"linear_attention","op":"rms_norm","roles":[],"inserted":false,"precision":{"weight":null,"activation":"bf16","output":"bf16"},"steps":["compute f32"],"group":58,"kernel":"residual_add_rms_norm_exact::residual_add_rms_norm_exact","family":"rms_norm","point":null,"compute":"memory","numerics":"bit_identical(residual_add_rms_norm_exact_microtest)","where":"device","state":null},{"id":"l4.gdn.qkvz","order":96,"layer":4,"block":"l4.gdn","block_type":"gdn","kind":"linear_attention","op":"linear:qkvz","roles":["qkvz"],"inserted":false,"precision":{"weight":"fp8/block128x128","activation":"bf16","output":"bf16"},"steps":["act bf16","weight fp8/block128x128->bf16","mma bf16*bf16","accumulate f32","scale f32"],"group":59,"kernel":"w8a16_gemm_pipelined_m32::w8a16_gemm_pipelined_m32","family":"w8a16_gemm","point":null,"compute":"tensor_core:mma.sync.m16n8k16.bf16","numerics":"reference","where":"device","state":null},{"id":"l4.gdn.ba","order":97,"layer":4,"block":"l4.gdn","block_type":"gdn","kind":"linear_attention","op":"linear:ba","roles":["ba"],"inserted":false,"precision":{"weight":"bf16","activation":"bf16","output":"f32"},"steps":["act bf16","weight bf16->bf16","mma bf16*bf16","accumulate f32","scale none"],"group":60,"kernel":"ssm_preprocess::dense_gemv_ba_gates","family":"gdn_ba_gates_gemv","point":{"gate":"softplus","decay":"head"},"compute":"cuda_core","numerics":"reference","where":"device","state":null},{"id":"l4.gdn.gates","order":98,"layer":4,"block":"l4.gdn","block_type":"gdn","kind":"linear_attention","op":"gdn_gates","roles":[],"inserted":false,"precision":{"weight":null,"activation":"f32","output":"f32,f32"},"steps":["compute f32"],"group":60,"kernel":"ssm_preprocess::dense_gemv_ba_gates","family":"gdn_ba_gates_gemv","point":{"gate":"softplus","decay":"head"},"compute":"cuda_core","numerics":"reference","where":"device","state":null},{"id":"l4.gdn.conv","order":99,"layer":4,"block":"l4.gdn","block_type":"gdn","kind":"linear_attention","op":"conv1d_update","roles":[],"inserted":false,"precision":{"weight":null,"activation":"bf16","output":"f32"},"steps":["state f32","compute f32"],"group":61,"kernel":"causal_conv1d::causal_conv1d_update_l2norm_f32","family":"causal_conv1d_l2norm","point":null,"compute":"cuda_core","numerics":"reference","where":"device","state":"conv:update"},{"id":"l4.gdn.conv_ckpt","order":100,"layer":4,"block":"l4.gdn","block_type":"gdn","kind":"linear_attention","op":"state_snapshot","roles":[],"inserted":false,"precision":{"weight":null,"activation":"f32","output":null},"steps":["state f32"],"group":61,"kernel":"causal_conv1d::causal_conv1d_update_l2norm_f32","family":"causal_conv1d_l2norm","point":null,"compute":"cuda_core","numerics":"reference","where":"device","state":"conv:snapshot"},{"id":"l4.gdn.l2","order":101,"layer":4,"block":"l4.gdn","block_type":"gdn","kind":"linear_attention","op":"l2_norm","roles":[],"inserted":false,"precision":{"weight":null,"activation":"f32","output":"f32"},"steps":["compute f32"],"group":61,"kernel":"causal_conv1d::causal_conv1d_update_l2norm_f32","family":"causal_conv1d_l2norm","point":null,"compute":"cuda_core","numerics":"reference","where":"device","state":null},{"id":"l4.gdn.recur","order":102,"layer":4,"block":"l4.gdn","block_type":"gdn","kind":"linear_attention","op":"gdn_recurrence","roles":[],"inserted":false,"precision":{"weight":null,"activation":"f32","output":"f32"},"steps":["state f32","compute f32"],"group":62,"kernel":"gated_delta_rule::gated_delta_rule_decode_f32","family":"gdn_recurrence","point":{"decay":"head","gate_act":"silu"},"compute":"cuda_core","numerics":"reference","where":"device","state":"h:update"},{"id":"l4.gdn.out_norm","order":103,"layer":4,"block":"l4.gdn","block_type":"gdn","kind":"linear_attention","op":"gated_rms_norm","roles":[],"inserted":false,"precision":{"weight":null,"activation":"f32","output":"bf16"},"steps":["compute f32"],"group":63,"kernel":"norm::gated_rms_norm_f32_input","family":"gated_rms_norm","point":null,"compute":"memory","numerics":"reference","where":"device","state":null},{"id":"l4.gdn.out","order":104,"layer":4,"block":"l4.gdn","block_type":"gdn","kind":"linear_attention","op":"linear:gdn_out","roles":["gdn_out"],"inserted":false,"precision":{"weight":"fp8/block128x128","activation":"bf16","output":"bf16"},"steps":["act bf16","weight fp8/block128x128->bf16","mma bf16*bf16","accumulate f32","scale f32"],"group":64,"kernel":"w8a16_gemm_pipelined_m32::w8a16_gemm_pipelined_m32","family":"w8a16_gemm","point":null,"compute":"tensor_core:mma.sync.m16n8k16.bf16","numerics":"reference","where":"device","state":null},{"id":"l4.gdn.add","order":105,"layer":4,"block":"l4.gdn","block_type":"gdn","kind":"linear_attention","op":"residual_add","roles":[],"inserted":false,"precision":{"weight":null,"activation":"bf16","output":"bf16"},"steps":["compute f32"],"group":65,"kernel":"norm::residual_add_rms_norm","family":"rms_norm","point":null,"compute":"memory","numerics":"reference","where":"device","state":null},{"id":"l4.moe_ffn.post_norm","order":106,"layer":4,"block":"l4.moe_ffn","block_type":"moe_ffn","kind":"linear_attention","op":"rms_norm","roles":[],"inserted":false,"precision":{"weight":null,"activation":"f32","output":"bf16"},"steps":["compute f32"],"group":65,"kernel":"norm::residual_add_rms_norm","family":"rms_norm","point":null,"compute":"memory","numerics":"reference","where":"device","state":null},{"id":"l4.moe_ffn.router","order":107,"layer":4,"block":"l4.moe_ffn","block_type":"moe_ffn","kind":"linear_attention","op":"router","roles":[],"inserted":false,"precision":{"weight":"bf16","activation":"bf16","output":"bf16"},"steps":["act bf16","weight bf16->bf16","mma bf16*bf16","accumulate f32","scale none"],"group":66,"kernel":"gemv::dense_gemv_bf16","family":"dense_bf16","point":null,"compute":"cuda_core","numerics":"reference","where":"device","state":null},{"id":"l4.moe_ffn.top_k","order":108,"layer":4,"block":"l4.moe_ffn","block_type":"moe_ffn","kind":"linear_attention","op":"top_k","roles":[],"inserted":false,"precision":{"weight":null,"activation":"bf16","output":"f32,i32"},"steps":["score f32"],"group":67,"kernel":"moe_topk::moe_topk_softmax","family":"moe_topk","point":null,"compute":"memory","numerics":"reference","where":"device","state":null},{"id":"l4.moe_ffn.experts_gate_up","order":109,"layer":4,"block":"l4.moe_ffn","block_type":"moe_ffn","kind":"linear_attention","op":"expert_gate_up","roles":[],"inserted":false,"precision":{"weight":"fp8/block128x128","activation":"bf16","output":"bf16"},"steps":["gather bf16","act bf16","weight fp8/block128x128->bf16","mma bf16*bf16","accumulate f32","scale f32"],"group":68,"kernel":"moe_shared_expert_fused_fp8::moe_expert_gate_up_shared_fp8","family":"moe_fp8_1row","point":null,"compute":"cuda_core","numerics":"reference","where":"device","state":null},{"id":"l4.moe_ffn.shared_gate_up","order":110,"layer":4,"block":"l4.moe_ffn","block_type":"moe_ffn","kind":"linear_attention","op":"linear:shared_gate_up","roles":["shared_gate_up"],"inserted":false,"precision":{"weight":"fp8/block128x128","activation":"bf16","output":"bf16"},"steps":["act bf16","weight fp8/block128x128->bf16","mma bf16*bf16","accumulate f32","scale f32"],"group":68,"kernel":"moe_shared_expert_fused_fp8::moe_expert_gate_up_shared_fp8","family":"moe_fp8_1row","point":null,"compute":"cuda_core","numerics":"reference","where":"device","state":null},{"id":"l4.moe_ffn.experts_act","order":111,"layer":4,"block":"l4.moe_ffn","block_type":"moe_ffn","kind":"linear_attention","op":"silu_mul","roles":[],"inserted":false,"precision":{"weight":null,"activation":"bf16","output":"f32"},"steps":["compute f32"],"group":69,"kernel":"moe_shared_expert_fused_fp8::moe_expert_silu_down_shared_fp8","family":"moe_fp8_1row","point":null,"compute":"cuda_core","numerics":"reference","where":"device","state":null},{"id":"l4.moe_ffn.experts_down","order":112,"layer":4,"block":"l4.moe_ffn","block_type":"moe_ffn","kind":"linear_attention","op":"expert_down","roles":[],"inserted":false,"precision":{"weight":"fp8/block128x128","activation":"f32","output":"bf16"},"steps":["act f32","weight fp8/block128x128->f32","mma f32*f32","accumulate f32","scale f32"],"group":69,"kernel":"moe_shared_expert_fused_fp8::moe_expert_silu_down_shared_fp8","family":"moe_fp8_1row","point":null,"compute":"cuda_core","numerics":"reference","where":"device","state":null},{"id":"l4.moe_ffn.shared_act","order":113,"layer":4,"block":"l4.moe_ffn","block_type":"moe_ffn","kind":"linear_attention","op":"silu_mul","roles":[],"inserted":false,"precision":{"weight":null,"activation":"bf16","output":"f32"},"steps":["compute f32"],"group":69,"kernel":"moe_shared_expert_fused_fp8::moe_expert_silu_down_shared_fp8","family":"moe_fp8_1row","point":null,"compute":"cuda_core","numerics":"reference","where":"device","state":null},{"id":"l4.moe_ffn.shared_down","order":114,"layer":4,"block":"l4.moe_ffn","block_type":"moe_ffn","kind":"linear_attention","op":"linear:shared_down","roles":["shared_down"],"inserted":false,"precision":{"weight":"fp8/block128x128","activation":"f32","output":"bf16"},"steps":["act f32","weight fp8/block128x128->f32","mma f32*f32","accumulate f32","scale f32"],"group":69,"kernel":"moe_shared_expert_fused_fp8::moe_expert_silu_down_shared_fp8","family":"moe_fp8_1row","point":null,"compute":"cuda_core","numerics":"reference","where":"device","state":null},{"id":"l4.moe_ffn.shared_gate","order":115,"layer":4,"block":"l4.moe_ffn","block_type":"moe_ffn","kind":"linear_attention","op":"linear:shared_gate","roles":["shared_gate"],"inserted":false,"precision":{"weight":"bf16","activation":"bf16","output":"f32"},"steps":["act bf16","weight bf16->bf16","mma bf16*bf16","accumulate f32","scale none"],"group":70,"kernel":"moe_expert_gemv::moe_weighted_sum_blend","family":"moe_blend","point":{"shared_gate":"sigmoid"},"compute":"cuda_core","numerics":"reference","where":"device","state":null},{"id":"l4.moe_ffn.blend","order":116,"layer":4,"block":"l4.moe_ffn","block_type":"moe_ffn","kind":"linear_attention","op":"blend","roles":[],"inserted":false,"precision":{"weight":null,"activation":"bf16","output":"bf16"},"steps":["scatter f32","combine f32"],"group":70,"kernel":"moe_expert_gemv::moe_weighted_sum_blend","family":"moe_blend","point":{"shared_gate":"sigmoid"},"compute":"cuda_core","numerics":"reference","where":"device","state":null},{"id":"l4.moe_ffn.add","order":117,"layer":4,"block":"l4.moe_ffn","block_type":"moe_ffn","kind":"linear_attention","op":"residual_add","roles":[],"inserted":false,"precision":{"weight":null,"activation":"bf16","output":"bf16"},"steps":["compute f32"],"group":71,"kernel":"residual_add_rms_norm_exact::residual_add_rms_norm_exact","family":"rms_norm","point":null,"compute":"memory","numerics":"bit_identical(residual_add_rms_norm_exact_microtest)","where":"device","state":null},{"id":"l5.gdn.input_norm","order":118,"layer":5,"block":"l5.gdn","block_type":"gdn","kind":"linear_attention","op":"rms_norm","roles":[],"inserted":false,"precision":{"weight":null,"activation":"bf16","output":"bf16"},"steps":["compute f32"],"group":71,"kernel":"residual_add_rms_norm_exact::residual_add_rms_norm_exact","family":"rms_norm","point":null,"compute":"memory","numerics":"bit_identical(residual_add_rms_norm_exact_microtest)","where":"device","state":null},{"id":"l5.gdn.qkvz","order":119,"layer":5,"block":"l5.gdn","block_type":"gdn","kind":"linear_attention","op":"linear:qkvz","roles":["qkvz"],"inserted":false,"precision":{"weight":"fp8/block128x128","activation":"bf16","output":"bf16"},"steps":["act bf16","weight fp8/block128x128->bf16","mma bf16*bf16","accumulate f32","scale f32"],"group":72,"kernel":"w8a16_gemm_pipelined_m32::w8a16_gemm_pipelined_m32","family":"w8a16_gemm","point":null,"compute":"tensor_core:mma.sync.m16n8k16.bf16","numerics":"reference","where":"device","state":null},{"id":"l5.gdn.ba","order":120,"layer":5,"block":"l5.gdn","block_type":"gdn","kind":"linear_attention","op":"linear:ba","roles":["ba"],"inserted":false,"precision":{"weight":"bf16","activation":"bf16","output":"f32"},"steps":["act bf16","weight bf16->bf16","mma bf16*bf16","accumulate f32","scale none"],"group":73,"kernel":"ssm_preprocess::dense_gemv_ba_gates","family":"gdn_ba_gates_gemv","point":{"gate":"softplus","decay":"head"},"compute":"cuda_core","numerics":"reference","where":"device","state":null},{"id":"l5.gdn.gates","order":121,"layer":5,"block":"l5.gdn","block_type":"gdn","kind":"linear_attention","op":"gdn_gates","roles":[],"inserted":false,"precision":{"weight":null,"activation":"f32","output":"f32,f32"},"steps":["compute f32"],"group":73,"kernel":"ssm_preprocess::dense_gemv_ba_gates","family":"gdn_ba_gates_gemv","point":{"gate":"softplus","decay":"head"},"compute":"cuda_core","numerics":"reference","where":"device","state":null},{"id":"l5.gdn.conv","order":122,"layer":5,"block":"l5.gdn","block_type":"gdn","kind":"linear_attention","op":"conv1d_update","roles":[],"inserted":false,"precision":{"weight":null,"activation":"bf16","output":"f32"},"steps":["state f32","compute f32"],"group":74,"kernel":"causal_conv1d::causal_conv1d_update_l2norm_f32","family":"causal_conv1d_l2norm","point":null,"compute":"cuda_core","numerics":"reference","where":"device","state":"conv:update"},{"id":"l5.gdn.conv_ckpt","order":123,"layer":5,"block":"l5.gdn","block_type":"gdn","kind":"linear_attention","op":"state_snapshot","roles":[],"inserted":false,"precision":{"weight":null,"activation":"f32","output":null},"steps":["state f32"],"group":74,"kernel":"causal_conv1d::causal_conv1d_update_l2norm_f32","family":"causal_conv1d_l2norm","point":null,"compute":"cuda_core","numerics":"reference","where":"device","state":"conv:snapshot"},{"id":"l5.gdn.l2","order":124,"layer":5,"block":"l5.gdn","block_type":"gdn","kind":"linear_attention","op":"l2_norm","roles":[],"inserted":false,"precision":{"weight":null,"activation":"f32","output":"f32"},"steps":["compute f32"],"group":74,"kernel":"causal_conv1d::causal_conv1d_update_l2norm_f32","family":"causal_conv1d_l2norm","point":null,"compute":"cuda_core","numerics":"reference","where":"device","state":null},{"id":"l5.gdn.recur","order":125,"layer":5,"block":"l5.gdn","block_type":"gdn","kind":"linear_attention","op":"gdn_recurrence","roles":[],"inserted":false,"precision":{"weight":null,"activation":"f32","output":"f32"},"steps":["state f32","compute f32"],"group":75,"kernel":"gated_delta_rule::gated_delta_rule_decode_f32","family":"gdn_recurrence","point":{"decay":"head","gate_act":"silu"},"compute":"cuda_core","numerics":"reference","where":"device","state":"h:update"},{"id":"l5.gdn.out_norm","order":126,"layer":5,"block":"l5.gdn","block_type":"gdn","kind":"linear_attention","op":"gated_rms_norm","roles":[],"inserted":false,"precision":{"weight":null,"activation":"f32","output":"bf16"},"steps":["compute f32"],"group":76,"kernel":"norm::gated_rms_norm_f32_input","family":"gated_rms_norm","point":null,"compute":"memory","numerics":"reference","where":"device","state":null},{"id":"l5.gdn.out","order":127,"layer":5,"block":"l5.gdn","block_type":"gdn","kind":"linear_attention","op":"linear:gdn_out","roles":["gdn_out"],"inserted":false,"precision":{"weight":"fp8/block128x128","activation":"bf16","output":"bf16"},"steps":["act bf16","weight fp8/block128x128->bf16","mma bf16*bf16","accumulate f32","scale f32"],"group":77,"kernel":"w8a16_gemm_pipelined_m32::w8a16_gemm_pipelined_m32","family":"w8a16_gemm","point":null,"compute":"tensor_core:mma.sync.m16n8k16.bf16","numerics":"reference","where":"device","state":null},{"id":"l5.gdn.add","order":128,"layer":5,"block":"l5.gdn","block_type":"gdn","kind":"linear_attention","op":"residual_add","roles":[],"inserted":false,"precision":{"weight":null,"activation":"bf16","output":"bf16"},"steps":["compute f32"],"group":78,"kernel":"norm::residual_add_rms_norm","family":"rms_norm","point":null,"compute":"memory","numerics":"reference","where":"device","state":null},{"id":"l5.moe_ffn.post_norm","order":129,"layer":5,"block":"l5.moe_ffn","block_type":"moe_ffn","kind":"linear_attention","op":"rms_norm","roles":[],"inserted":false,"precision":{"weight":null,"activation":"f32","output":"bf16"},"steps":["compute f32"],"group":78,"kernel":"norm::residual_add_rms_norm","family":"rms_norm","point":null,"compute":"memory","numerics":"reference","where":"device","state":null},{"id":"l5.moe_ffn.router","order":130,"layer":5,"block":"l5.moe_ffn","block_type":"moe_ffn","kind":"linear_attention","op":"router","roles":[],"inserted":false,"precision":{"weight":"bf16","activation":"bf16","output":"bf16"},"steps":["act bf16","weight bf16->bf16","mma bf16*bf16","accumulate f32","scale none"],"group":79,"kernel":"gemv::dense_gemv_bf16","family":"dense_bf16","point":null,"compute":"cuda_core","numerics":"reference","where":"device","state":null},{"id":"l5.moe_ffn.top_k","order":131,"layer":5,"block":"l5.moe_ffn","block_type":"moe_ffn","kind":"linear_attention","op":"top_k","roles":[],"inserted":false,"precision":{"weight":null,"activation":"bf16","output":"f32,i32"},"steps":["score f32"],"group":80,"kernel":"moe_topk::moe_topk_softmax","family":"moe_topk","point":null,"compute":"memory","numerics":"reference","where":"device","state":null},{"id":"l5.moe_ffn.experts_gate_up","order":132,"layer":5,"block":"l5.moe_ffn","block_type":"moe_ffn","kind":"linear_attention","op":"expert_gate_up","roles":[],"inserted":false,"precision":{"weight":"fp8/block128x128","activation":"bf16","output":"bf16"},"steps":["gather bf16","act bf16","weight fp8/block128x128->bf16","mma bf16*bf16","accumulate f32","scale f32"],"group":81,"kernel":"moe_shared_expert_fused_fp8::moe_expert_gate_up_shared_fp8","family":"moe_fp8_1row","point":null,"compute":"cuda_core","numerics":"reference","where":"device","state":null},{"id":"l5.moe_ffn.shared_gate_up","order":133,"layer":5,"block":"l5.moe_ffn","block_type":"moe_ffn","kind":"linear_attention","op":"linear:shared_gate_up","roles":["shared_gate_up"],"inserted":false,"precision":{"weight":"fp8/block128x128","activation":"bf16","output":"bf16"},"steps":["act bf16","weight fp8/block128x128->bf16","mma bf16*bf16","accumulate f32","scale f32"],"group":81,"kernel":"moe_shared_expert_fused_fp8::moe_expert_gate_up_shared_fp8","family":"moe_fp8_1row","point":null,"compute":"cuda_core","numerics":"reference","where":"device","state":null},{"id":"l5.moe_ffn.experts_act","order":134,"layer":5,"block":"l5.moe_ffn","block_type":"moe_ffn","kind":"linear_attention","op":"silu_mul","roles":[],"inserted":false,"precision":{"weight":null,"activation":"bf16","output":"f32"},"steps":["compute f32"],"group":82,"kernel":"moe_shared_expert_fused_fp8::moe_expert_silu_down_shared_fp8","family":"moe_fp8_1row","point":null,"compute":"cuda_core","numerics":"reference","where":"device","state":null},{"id":"l5.moe_ffn.experts_down","order":135,"layer":5,"block":"l5.moe_ffn","block_type":"moe_ffn","kind":"linear_attention","op":"expert_down","roles":[],"inserted":false,"precision":{"weight":"fp8/block128x128","activation":"f32","output":"bf16"},"steps":["act f32","weight fp8/block128x128->f32","mma f32*f32","accumulate f32","scale f32"],"group":82,"kernel":"moe_shared_expert_fused_fp8::moe_expert_silu_down_shared_fp8","family":"moe_fp8_1row","point":null,"compute":"cuda_core","numerics":"reference","where":"device","state":null},{"id":"l5.moe_ffn.shared_act","order":136,"layer":5,"block":"l5.moe_ffn","block_type":"moe_ffn","kind":"linear_attention","op":"silu_mul","roles":[],"inserted":false,"precision":{"weight":null,"activation":"bf16","output":"f32"},"steps":["compute f32"],"group":82,"kernel":"moe_shared_expert_fused_fp8::moe_expert_silu_down_shared_fp8","family":"moe_fp8_1row","point":null,"compute":"cuda_core","numerics":"reference","where":"device","state":null},{"id":"l5.moe_ffn.shared_down","order":137,"layer":5,"block":"l5.moe_ffn","block_type":"moe_ffn","kind":"linear_attention","op":"linear:shared_down","roles":["shared_down"],"inserted":false,"precision":{"weight":"fp8/block128x128","activation":"f32","output":"bf16"},"steps":["act f32","weight fp8/block128x128->f32","mma f32*f32","accumulate f32","scale f32"],"group":82,"kernel":"moe_shared_expert_fused_fp8::moe_expert_silu_down_shared_fp8","family":"moe_fp8_1row","point":null,"compute":"cuda_core","numerics":"reference","where":"device","state":null},{"id":"l5.moe_ffn.shared_gate","order":138,"layer":5,"block":"l5.moe_ffn","block_type":"moe_ffn","kind":"linear_attention","op":"linear:shared_gate","roles":["shared_gate"],"inserted":false,"precision":{"weight":"bf16","activation":"bf16","output":"f32"},"steps":["act bf16","weight bf16->bf16","mma bf16*bf16","accumulate f32","scale none"],"group":83,"kernel":"moe_expert_gemv::moe_weighted_sum_blend","family":"moe_blend","point":{"shared_gate":"sigmoid"},"compute":"cuda_core","numerics":"reference","where":"device","state":null},{"id":"l5.moe_ffn.blend","order":139,"layer":5,"block":"l5.moe_ffn","block_type":"moe_ffn","kind":"linear_attention","op":"blend","roles":[],"inserted":false,"precision":{"weight":null,"activation":"bf16","output":"bf16"},"steps":["scatter f32","combine f32"],"group":83,"kernel":"moe_expert_gemv::moe_weighted_sum_blend","family":"moe_blend","point":{"shared_gate":"sigmoid"},"compute":"cuda_core","numerics":"reference","where":"device","state":null},{"id":"l5.moe_ffn.add","order":140,"layer":5,"block":"l5.moe_ffn","block_type":"moe_ffn","kind":"linear_attention","op":"residual_add","roles":[],"inserted":false,"precision":{"weight":null,"activation":"bf16","output":"bf16"},"steps":["compute f32"],"group":84,"kernel":"residual_add_rms_norm_exact::residual_add_rms_norm_exact","family":"rms_norm","point":null,"compute":"memory","numerics":"bit_identical(residual_add_rms_norm_exact_microtest)","where":"device","state":null},{"id":"l6.gdn.input_norm","order":141,"layer":6,"block":"l6.gdn","block_type":"gdn","kind":"linear_attention","op":"rms_norm","roles":[],"inserted":false,"precision":{"weight":null,"activation":"bf16","output":"bf16"},"steps":["compute f32"],"group":84,"kernel":"residual_add_rms_norm_exact::residual_add_rms_norm_exact","family":"rms_norm","point":null,"compute":"memory","numerics":"bit_identical(residual_add_rms_norm_exact_microtest)","where":"device","state":null},{"id":"l6.gdn.qkvz","order":142,"layer":6,"block":"l6.gdn","block_type":"gdn","kind":"linear_attention","op":"linear:qkvz","roles":["qkvz"],"inserted":false,"precision":{"weight":"fp8/block128x128","activation":"bf16","output":"bf16"},"steps":["act bf16","weight fp8/block128x128->bf16","mma bf16*bf16","accumulate f32","scale f32"],"group":85,"kernel":"w8a16_gemm_pipelined_m32::w8a16_gemm_pipelined_m32","family":"w8a16_gemm","point":null,"compute":"tensor_core:mma.sync.m16n8k16.bf16","numerics":"reference","where":"device","state":null},{"id":"l6.gdn.ba","order":143,"layer":6,"block":"l6.gdn","block_type":"gdn","kind":"linear_attention","op":"linear:ba","roles":["ba"],"inserted":false,"precision":{"weight":"bf16","activation":"bf16","output":"f32"},"steps":["act bf16","weight bf16->bf16","mma bf16*bf16","accumulate f32","scale none"],"group":86,"kernel":"ssm_preprocess::dense_gemv_ba_gates","family":"gdn_ba_gates_gemv","point":{"gate":"softplus","decay":"head"},"compute":"cuda_core","numerics":"reference","where":"device","state":null},{"id":"l6.gdn.gates","order":144,"layer":6,"block":"l6.gdn","block_type":"gdn","kind":"linear_attention","op":"gdn_gates","roles":[],"inserted":false,"precision":{"weight":null,"activation":"f32","output":"f32,f32"},"steps":["compute f32"],"group":86,"kernel":"ssm_preprocess::dense_gemv_ba_gates","family":"gdn_ba_gates_gemv","point":{"gate":"softplus","decay":"head"},"compute":"cuda_core","numerics":"reference","where":"device","state":null},{"id":"l6.gdn.conv","order":145,"layer":6,"block":"l6.gdn","block_type":"gdn","kind":"linear_attention","op":"conv1d_update","roles":[],"inserted":false,"precision":{"weight":null,"activation":"bf16","output":"f32"},"steps":["state f32","compute f32"],"group":87,"kernel":"causal_conv1d::causal_conv1d_update_l2norm_f32","family":"causal_conv1d_l2norm","point":null,"compute":"cuda_core","numerics":"reference","where":"device","state":"conv:update"},{"id":"l6.gdn.conv_ckpt","order":146,"layer":6,"block":"l6.gdn","block_type":"gdn","kind":"linear_attention","op":"state_snapshot","roles":[],"inserted":false,"precision":{"weight":null,"activation":"f32","output":null},"steps":["state f32"],"group":87,"kernel":"causal_conv1d::causal_conv1d_update_l2norm_f32","family":"causal_conv1d_l2norm","point":null,"compute":"cuda_core","numerics":"reference","where":"device","state":"conv:snapshot"},{"id":"l6.gdn.l2","order":147,"layer":6,"block":"l6.gdn","block_type":"gdn","kind":"linear_attention","op":"l2_norm","roles":[],"inserted":false,"precision":{"weight":null,"activation":"f32","output":"f32"},"steps":["compute f32"],"group":87,"kernel":"causal_conv1d::causal_conv1d_update_l2norm_f32","family":"causal_conv1d_l2norm","point":null,"compute":"cuda_core","numerics":"reference","where":"device","state":null},{"id":"l6.gdn.recur","order":148,"layer":6,"block":"l6.gdn","block_type":"gdn","kind":"linear_attention","op":"gdn_recurrence","roles":[],"inserted":false,"precision":{"weight":null,"activation":"f32","output":"f32"},"steps":["state f32","compute f32"],"group":88,"kernel":"gated_delta_rule::gated_delta_rule_decode_f32","family":"gdn_recurrence","point":{"decay":"head","gate_act":"silu"},"compute":"cuda_core","numerics":"reference","where":"device","state":"h:update"},{"id":"l6.gdn.out_norm","order":149,"layer":6,"block":"l6.gdn","block_type":"gdn","kind":"linear_attention","op":"gated_rms_norm","roles":[],"inserted":false,"precision":{"weight":null,"activation":"f32","output":"bf16"},"steps":["compute f32"],"group":89,"kernel":"norm::gated_rms_norm_f32_input","family":"gated_rms_norm","point":null,"compute":"memory","numerics":"reference","where":"device","state":null},{"id":"l6.gdn.out","order":150,"layer":6,"block":"l6.gdn","block_type":"gdn","kind":"linear_attention","op":"linear:gdn_out","roles":["gdn_out"],"inserted":false,"precision":{"weight":"fp8/block128x128","activation":"bf16","output":"bf16"},"steps":["act bf16","weight fp8/block128x128->bf16","mma bf16*bf16","accumulate f32","scale f32"],"group":90,"kernel":"w8a16_gemm_pipelined_m32::w8a16_gemm_pipelined_m32","family":"w8a16_gemm","point":null,"compute":"tensor_core:mma.sync.m16n8k16.bf16","numerics":"reference","where":"device","state":null},{"id":"l6.gdn.add","order":151,"layer":6,"block":"l6.gdn","block_type":"gdn","kind":"linear_attention","op":"residual_add","roles":[],"inserted":false,"precision":{"weight":null,"activation":"bf16","output":"bf16"},"steps":["compute f32"],"group":91,"kernel":"norm::residual_add_rms_norm","family":"rms_norm","point":null,"compute":"memory","numerics":"reference","where":"device","state":null},{"id":"l6.moe_ffn.post_norm","order":152,"layer":6,"block":"l6.moe_ffn","block_type":"moe_ffn","kind":"linear_attention","op":"rms_norm","roles":[],"inserted":false,"precision":{"weight":null,"activation":"f32","output":"bf16"},"steps":["compute f32"],"group":91,"kernel":"norm::residual_add_rms_norm","family":"rms_norm","point":null,"compute":"memory","numerics":"reference","where":"device","state":null},{"id":"l6.moe_ffn.router","order":153,"layer":6,"block":"l6.moe_ffn","block_type":"moe_ffn","kind":"linear_attention","op":"router","roles":[],"inserted":false,"precision":{"weight":"bf16","activation":"bf16","output":"bf16"},"steps":["act bf16","weight bf16->bf16","mma bf16*bf16","accumulate f32","scale none"],"group":92,"kernel":"gemv::dense_gemv_bf16","family":"dense_bf16","point":null,"compute":"cuda_core","numerics":"reference","where":"device","state":null},{"id":"l6.moe_ffn.top_k","order":154,"layer":6,"block":"l6.moe_ffn","block_type":"moe_ffn","kind":"linear_attention","op":"top_k","roles":[],"inserted":false,"precision":{"weight":null,"activation":"bf16","output":"f32,i32"},"steps":["score f32"],"group":93,"kernel":"moe_topk::moe_topk_softmax","family":"moe_topk","point":null,"compute":"memory","numerics":"reference","where":"device","state":null},{"id":"l6.moe_ffn.experts_gate_up","order":155,"layer":6,"block":"l6.moe_ffn","block_type":"moe_ffn","kind":"linear_attention","op":"expert_gate_up","roles":[],"inserted":false,"precision":{"weight":"fp8/block128x128","activation":"bf16","output":"bf16"},"steps":["gather bf16","act bf16","weight fp8/block128x128->bf16","mma bf16*bf16","accumulate f32","scale f32"],"group":94,"kernel":"moe_shared_expert_fused_fp8::moe_expert_gate_up_shared_fp8","family":"moe_fp8_1row","point":null,"compute":"cuda_core","numerics":"reference","where":"device","state":null},{"id":"l6.moe_ffn.shared_gate_up","order":156,"layer":6,"block":"l6.moe_ffn","block_type":"moe_ffn","kind":"linear_attention","op":"linear:shared_gate_up","roles":["shared_gate_up"],"inserted":false,"precision":{"weight":"fp8/block128x128","activation":"bf16","output":"bf16"},"steps":["act bf16","weight fp8/block128x128->bf16","mma bf16*bf16","accumulate f32","scale f32"],"group":94,"kernel":"moe_shared_expert_fused_fp8::moe_expert_gate_up_shared_fp8","family":"moe_fp8_1row","point":null,"compute":"cuda_core","numerics":"reference","where":"device","state":null},{"id":"l6.moe_ffn.experts_act","order":157,"layer":6,"block":"l6.moe_ffn","block_type":"moe_ffn","kind":"linear_attention","op":"silu_mul","roles":[],"inserted":false,"precision":{"weight":null,"activation":"bf16","output":"f32"},"steps":["compute f32"],"group":95,"kernel":"moe_shared_expert_fused_fp8::moe_expert_silu_down_shared_fp8","family":"moe_fp8_1row","point":null,"compute":"cuda_core","numerics":"reference","where":"device","state":null},{"id":"l6.moe_ffn.experts_down","order":158,"layer":6,"block":"l6.moe_ffn","block_type":"moe_ffn","kind":"linear_attention","op":"expert_down","roles":[],"inserted":false,"precision":{"weight":"fp8/block128x128","activation":"f32","output":"bf16"},"steps":["act f32","weight fp8/block128x128->f32","mma f32*f32","accumulate f32","scale f32"],"group":95,"kernel":"moe_shared_expert_fused_fp8::moe_expert_silu_down_shared_fp8","family":"moe_fp8_1row","point":null,"compute":"cuda_core","numerics":"reference","where":"device","state":null},{"id":"l6.moe_ffn.shared_act","order":159,"layer":6,"block":"l6.moe_ffn","block_type":"moe_ffn","kind":"linear_attention","op":"silu_mul","roles":[],"inserted":false,"precision":{"weight":null,"activation":"bf16","output":"f32"},"steps":["compute f32"],"group":95,"kernel":"moe_shared_expert_fused_fp8::moe_expert_silu_down_shared_fp8","family":"moe_fp8_1row","point":null,"compute":"cuda_core","numerics":"reference","where":"device","state":null},{"id":"l6.moe_ffn.shared_down","order":160,"layer":6,"block":"l6.moe_ffn","block_type":"moe_ffn","kind":"linear_attention","op":"linear:shared_down","roles":["shared_down"],"inserted":false,"precision":{"weight":"fp8/block128x128","activation":"f32","output":"bf16"},"steps":["act f32","weight fp8/block128x128->f32","mma f32*f32","accumulate f32","scale f32"],"group":95,"kernel":"moe_shared_expert_fused_fp8::moe_expert_silu_down_shared_fp8","family":"moe_fp8_1row","point":null,"compute":"cuda_core","numerics":"reference","where":"device","state":null},{"id":"l6.moe_ffn.shared_gate","order":161,"layer":6,"block":"l6.moe_ffn","block_type":"moe_ffn","kind":"linear_attention","op":"linear:shared_gate","roles":["shared_gate"],"inserted":false,"precision":{"weight":"bf16","activation":"bf16","output":"f32"},"steps":["act bf16","weight bf16->bf16","mma bf16*bf16","accumulate f32","scale none"],"group":96,"kernel":"moe_expert_gemv::moe_weighted_sum_blend","family":"moe_blend","point":{"shared_gate":"sigmoid"},"compute":"cuda_core","numerics":"reference","where":"device","state":null},{"id":"l6.moe_ffn.blend","order":162,"layer":6,"block":"l6.moe_ffn","block_type":"moe_ffn","kind":"linear_attention","op":"blend","roles":[],"inserted":false,"precision":{"weight":null,"activation":"bf16","output":"bf16"},"steps":["scatter f32","combine f32"],"group":96,"kernel":"moe_expert_gemv::moe_weighted_sum_blend","family":"moe_blend","point":{"shared_gate":"sigmoid"},"compute":"cuda_core","numerics":"reference","where":"device","state":null},{"id":"l6.moe_ffn.add","order":163,"layer":6,"block":"l6.moe_ffn","block_type":"moe_ffn","kind":"linear_attention","op":"residual_add","roles":[],"inserted":false,"precision":{"weight":null,"activation":"bf16","output":"bf16"},"steps":["compute f32"],"group":97,"kernel":"residual_add_rms_norm_exact::residual_add_rms_norm_exact","family":"rms_norm","point":null,"compute":"memory","numerics":"bit_identical(residual_add_rms_norm_exact_microtest)","where":"device","state":null},{"id":"l7.attn.input_norm","order":164,"layer":7,"block":"l7.attn","block_type":"attn","kind":"full_attention","op":"rms_norm","roles":[],"inserted":false,"precision":{"weight":null,"activation":"bf16","output":"bf16"},"steps":["compute f32"],"group":97,"kernel":"residual_add_rms_norm_exact::residual_add_rms_norm_exact","family":"rms_norm","point":null,"compute":"memory","numerics":"bit_identical(residual_add_rms_norm_exact_microtest)","where":"device","state":null},{"id":"l7.attn.q","order":165,"layer":7,"block":"l7.attn","block_type":"attn","kind":"full_attention","op":"linear:q","roles":["q"],"inserted":false,"precision":{"weight":"fp8/block128x128","activation":"bf16","output":"bf16"},"steps":["act bf16","weight fp8/block128x128->bf16","mma bf16*bf16","accumulate f32","scale f32"],"group":98,"kernel":"w8a16_gemm_pipelined_m32::w8a16_gemm_pipelined_m32","family":"w8a16_gemm","point":null,"compute":"tensor_core:mma.sync.m16n8k16.bf16","numerics":"reference","where":"device","state":null},{"id":"l7.attn.q_split","order":166,"layer":7,"block":"l7.attn","block_type":"attn","kind":"full_attention","op":"split","roles":[],"inserted":false,"precision":{"weight":null,"activation":"bf16","output":"bf16,bf16"},"steps":["move bf16"],"group":99,"kernel":"ssm_preprocess::deinterleave_qg","family":"deinterleave_qg","point":{"layout":"interleaved"},"compute":"memory","numerics":"reference","where":"device","state":null},{"id":"l7.attn.k","order":167,"layer":7,"block":"l7.attn","block_type":"attn","kind":"full_attention","op":"linear:k","roles":["k"],"inserted":false,"precision":{"weight":"fp8/block128x128","activation":"bf16","output":"bf16"},"steps":["act bf16","weight fp8/block128x128->bf16","mma bf16*bf16","accumulate f32","scale f32"],"group":100,"kernel":"w8a16_gemm_pipelined_m32::w8a16_gemm_pipelined_m32","family":"w8a16_gemm","point":null,"compute":"tensor_core:mma.sync.m16n8k16.bf16","numerics":"reference","where":"device","state":null},{"id":"l7.attn.v","order":168,"layer":7,"block":"l7.attn","block_type":"attn","kind":"full_attention","op":"linear:v","roles":["v"],"inserted":false,"precision":{"weight":"fp8/block128x128","activation":"bf16","output":"bf16"},"steps":["act bf16","weight fp8/block128x128->bf16","mma bf16*bf16","accumulate f32","scale f32"],"group":101,"kernel":"w8a16_gemm_pipelined_m32::w8a16_gemm_pipelined_m32","family":"w8a16_gemm","point":null,"compute":"tensor_core:mma.sync.m16n8k16.bf16","numerics":"reference","where":"device","state":null},{"id":"l7.attn.q_norm","order":169,"layer":7,"block":"l7.attn","block_type":"attn","kind":"full_attention","op":"qk_norm","roles":[],"inserted":false,"precision":{"weight":null,"activation":"bf16","output":"bf16"},"steps":["compute f32"],"group":102,"kernel":"norm::rms_norm","family":"rms_norm","point":null,"compute":"memory","numerics":"reference","where":"device","state":null},{"id":"l7.attn.k_norm","order":170,"layer":7,"block":"l7.attn","block_type":"attn","kind":"full_attention","op":"qk_norm","roles":[],"inserted":false,"precision":{"weight":null,"activation":"bf16","output":"bf16"},"steps":["compute f32"],"group":103,"kernel":"norm::rms_norm","family":"rms_norm","point":null,"compute":"memory","numerics":"reference","where":"device","state":null},{"id":"l7.attn.rope","order":171,"layer":7,"block":"l7.attn","block_type":"attn","kind":"full_attention","op":"rope","roles":[],"inserted":false,"precision":{"weight":null,"activation":"bf16","output":"bf16,bf16"},"steps":["compute f32"],"group":104,"kernel":"rope_mrope_interleaved::rope_forward_mrope_interleaved","family":"rope","point":null,"compute":"memory","numerics":"reference","where":"device","state":null},{"id":"l7.attn.kv_write","order":172,"layer":7,"block":"l7.attn","block_type":"attn","kind":"full_attention","op":"kv_write","roles":[],"inserted":false,"precision":{"weight":null,"activation":"bf16","output":null},"steps":["cache bf16"],"group":105,"kernel":"reshape_and_cache::reshape_and_cache_flash","family":"kv_write","point":null,"compute":"memory","numerics":"reference","where":"device","state":"k:write,v:write"},{"id":"l7.attn.attend","order":173,"layer":7,"block":"l7.attn","block_type":"attn","kind":"full_attention","op":"paged_attention","roles":[],"inserted":false,"precision":{"weight":null,"activation":"bf16","output":"bf16"},"steps":["cache bf16","scores f32","softmax f32","accumulate f32"],"group":106,"kernel":"paged_decode::paged_decode_attn","family":"paged_decode_attn","point":null,"compute":"cuda_core","numerics":"reference","where":"device","state":"k:read,v:read"},{"id":"l7.attn.gate_mul","order":174,"layer":7,"block":"l7.attn","block_type":"attn","kind":"full_attention","op":"sigmoid_gate_mul","roles":[],"inserted":false,"precision":{"weight":null,"activation":"bf16","output":"bf16"},"steps":["compute f32"],"group":107,"kernel":"residual_add::sigmoid_gate_mul","family":"sigmoid_gate_mul","point":null,"compute":"memory","numerics":"reference","where":"device","state":null},{"id":"l7.attn.o","order":175,"layer":7,"block":"l7.attn","block_type":"attn","kind":"full_attention","op":"linear:o","roles":["o"],"inserted":false,"precision":{"weight":"fp8/block128x128","activation":"bf16","output":"bf16"},"steps":["act bf16","weight fp8/block128x128->bf16","mma bf16*bf16","accumulate f32","scale f32"],"group":108,"kernel":"w8a16_gemm_pipelined_m32::w8a16_gemm_pipelined_m32","family":"w8a16_gemm","point":null,"compute":"tensor_core:mma.sync.m16n8k16.bf16","numerics":"reference","where":"device","state":null},{"id":"l7.attn.add","order":176,"layer":7,"block":"l7.attn","block_type":"attn","kind":"full_attention","op":"residual_add","roles":[],"inserted":false,"precision":{"weight":null,"activation":"bf16","output":"bf16"},"steps":["compute f32"],"group":109,"kernel":"norm::residual_add_rms_norm","family":"rms_norm","point":null,"compute":"memory","numerics":"reference","where":"device","state":null},{"id":"l7.moe_ffn.post_norm","order":177,"layer":7,"block":"l7.moe_ffn","block_type":"moe_ffn","kind":"full_attention","op":"rms_norm","roles":[],"inserted":false,"precision":{"weight":null,"activation":"f32","output":"bf16"},"steps":["compute f32"],"group":109,"kernel":"norm::residual_add_rms_norm","family":"rms_norm","point":null,"compute":"memory","numerics":"reference","where":"device","state":null},{"id":"l7.moe_ffn.router","order":178,"layer":7,"block":"l7.moe_ffn","block_type":"moe_ffn","kind":"full_attention","op":"router","roles":[],"inserted":false,"precision":{"weight":"bf16","activation":"bf16","output":"bf16"},"steps":["act bf16","weight bf16->bf16","mma bf16*bf16","accumulate f32","scale none"],"group":110,"kernel":"gemv::dense_gemv_bf16","family":"dense_bf16","point":null,"compute":"cuda_core","numerics":"reference","where":"device","state":null},{"id":"l7.moe_ffn.top_k","order":179,"layer":7,"block":"l7.moe_ffn","block_type":"moe_ffn","kind":"full_attention","op":"top_k","roles":[],"inserted":false,"precision":{"weight":null,"activation":"bf16","output":"f32,i32"},"steps":["score f32"],"group":111,"kernel":"moe_topk::moe_topk_softmax","family":"moe_topk","point":null,"compute":"memory","numerics":"reference","where":"device","state":null},{"id":"l7.moe_ffn.experts_gate_up","order":180,"layer":7,"block":"l7.moe_ffn","block_type":"moe_ffn","kind":"full_attention","op":"expert_gate_up","roles":[],"inserted":false,"precision":{"weight":"fp8/block128x128","activation":"bf16","output":"bf16"},"steps":["gather bf16","act bf16","weight fp8/block128x128->bf16","mma bf16*bf16","accumulate f32","scale f32"],"group":112,"kernel":"moe_shared_expert_fused_fp8::moe_expert_gate_up_shared_fp8","family":"moe_fp8_1row","point":null,"compute":"cuda_core","numerics":"reference","where":"device","state":null},{"id":"l7.moe_ffn.shared_gate_up","order":181,"layer":7,"block":"l7.moe_ffn","block_type":"moe_ffn","kind":"full_attention","op":"linear:shared_gate_up","roles":["shared_gate_up"],"inserted":false,"precision":{"weight":"fp8/block128x128","activation":"bf16","output":"bf16"},"steps":["act bf16","weight fp8/block128x128->bf16","mma bf16*bf16","accumulate f32","scale f32"],"group":112,"kernel":"moe_shared_expert_fused_fp8::moe_expert_gate_up_shared_fp8","family":"moe_fp8_1row","point":null,"compute":"cuda_core","numerics":"reference","where":"device","state":null},{"id":"l7.moe_ffn.experts_act","order":182,"layer":7,"block":"l7.moe_ffn","block_type":"moe_ffn","kind":"full_attention","op":"silu_mul","roles":[],"inserted":false,"precision":{"weight":null,"activation":"bf16","output":"f32"},"steps":["compute f32"],"group":113,"kernel":"moe_shared_expert_fused_fp8::moe_expert_silu_down_shared_fp8","family":"moe_fp8_1row","point":null,"compute":"cuda_core","numerics":"reference","where":"device","state":null},{"id":"l7.moe_ffn.experts_down","order":183,"layer":7,"block":"l7.moe_ffn","block_type":"moe_ffn","kind":"full_attention","op":"expert_down","roles":[],"inserted":false,"precision":{"weight":"fp8/block128x128","activation":"f32","output":"bf16"},"steps":["act f32","weight fp8/block128x128->f32","mma f32*f32","accumulate f32","scale f32"],"group":113,"kernel":"moe_shared_expert_fused_fp8::moe_expert_silu_down_shared_fp8","family":"moe_fp8_1row","point":null,"compute":"cuda_core","numerics":"reference","where":"device","state":null},{"id":"l7.moe_ffn.shared_act","order":184,"layer":7,"block":"l7.moe_ffn","block_type":"moe_ffn","kind":"full_attention","op":"silu_mul","roles":[],"inserted":false,"precision":{"weight":null,"activation":"bf16","output":"f32"},"steps":["compute f32"],"group":113,"kernel":"moe_shared_expert_fused_fp8::moe_expert_silu_down_shared_fp8","family":"moe_fp8_1row","point":null,"compute":"cuda_core","numerics":"reference","where":"device","state":null},{"id":"l7.moe_ffn.shared_down","order":185,"layer":7,"block":"l7.moe_ffn","block_type":"moe_ffn","kind":"full_attention","op":"linear:shared_down","roles":["shared_down"],"inserted":false,"precision":{"weight":"fp8/block128x128","activation":"f32","output":"bf16"},"steps":["act f32","weight fp8/block128x128->f32","mma f32*f32","accumulate f32","scale f32"],"group":113,"kernel":"moe_shared_expert_fused_fp8::moe_expert_silu_down_shared_fp8","family":"moe_fp8_1row","point":null,"compute":"cuda_core","numerics":"reference","where":"device","state":null},{"id":"l7.moe_ffn.shared_gate","order":186,"layer":7,"block":"l7.moe_ffn","block_type":"moe_ffn","kind":"full_attention","op":"linear:shared_gate","roles":["shared_gate"],"inserted":false,"precision":{"weight":"bf16","activation":"bf16","output":"f32"},"steps":["act bf16","weight bf16->bf16","mma bf16*bf16","accumulate f32","scale none"],"group":114,"kernel":"moe_expert_gemv::moe_weighted_sum_blend","family":"moe_blend","point":{"shared_gate":"sigmoid"},"compute":"cuda_core","numerics":"reference","where":"device","state":null},{"id":"l7.moe_ffn.blend","order":187,"layer":7,"block":"l7.moe_ffn","block_type":"moe_ffn","kind":"full_attention","op":"blend","roles":[],"inserted":false,"precision":{"weight":null,"activation":"bf16","output":"bf16"},"steps":["scatter f32","combine f32"],"group":114,"kernel":"moe_expert_gemv::moe_weighted_sum_blend","family":"moe_blend","point":{"shared_gate":"sigmoid"},"compute":"cuda_core","numerics":"reference","where":"device","state":null},{"id":"l7.moe_ffn.add","order":188,"layer":7,"block":"l7.moe_ffn","block_type":"moe_ffn","kind":"full_attention","op":"residual_add","roles":[],"inserted":false,"precision":{"weight":null,"activation":"bf16","output":"bf16"},"steps":["compute f32"],"group":115,"kernel":"residual_add_rms_norm_exact::residual_add_rms_norm_exact","family":"rms_norm","point":null,"compute":"memory","numerics":"bit_identical(residual_add_rms_norm_exact_microtest)","where":"device","state":null},{"id":"l8.gdn.input_norm","order":189,"layer":8,"block":"l8.gdn","block_type":"gdn","kind":"linear_attention","op":"rms_norm","roles":[],"inserted":false,"precision":{"weight":null,"activation":"bf16","output":"bf16"},"steps":["compute f32"],"group":115,"kernel":"residual_add_rms_norm_exact::residual_add_rms_norm_exact","family":"rms_norm","point":null,"compute":"memory","numerics":"bit_identical(residual_add_rms_norm_exact_microtest)","where":"device","state":null},{"id":"l8.gdn.qkvz","order":190,"layer":8,"block":"l8.gdn","block_type":"gdn","kind":"linear_attention","op":"linear:qkvz","roles":["qkvz"],"inserted":false,"precision":{"weight":"fp8/block128x128","activation":"bf16","output":"bf16"},"steps":["act bf16","weight fp8/block128x128->bf16","mma bf16*bf16","accumulate f32","scale f32"],"group":116,"kernel":"w8a16_gemm_pipelined_m32::w8a16_gemm_pipelined_m32","family":"w8a16_gemm","point":null,"compute":"tensor_core:mma.sync.m16n8k16.bf16","numerics":"reference","where":"device","state":null},{"id":"l8.gdn.ba","order":191,"layer":8,"block":"l8.gdn","block_type":"gdn","kind":"linear_attention","op":"linear:ba","roles":["ba"],"inserted":false,"precision":{"weight":"bf16","activation":"bf16","output":"f32"},"steps":["act bf16","weight bf16->bf16","mma bf16*bf16","accumulate f32","scale none"],"group":117,"kernel":"ssm_preprocess::dense_gemv_ba_gates","family":"gdn_ba_gates_gemv","point":{"gate":"softplus","decay":"head"},"compute":"cuda_core","numerics":"reference","where":"device","state":null},{"id":"l8.gdn.gates","order":192,"layer":8,"block":"l8.gdn","block_type":"gdn","kind":"linear_attention","op":"gdn_gates","roles":[],"inserted":false,"precision":{"weight":null,"activation":"f32","output":"f32,f32"},"steps":["compute f32"],"group":117,"kernel":"ssm_preprocess::dense_gemv_ba_gates","family":"gdn_ba_gates_gemv","point":{"gate":"softplus","decay":"head"},"compute":"cuda_core","numerics":"reference","where":"device","state":null},{"id":"l8.gdn.conv","order":193,"layer":8,"block":"l8.gdn","block_type":"gdn","kind":"linear_attention","op":"conv1d_update","roles":[],"inserted":false,"precision":{"weight":null,"activation":"bf16","output":"f32"},"steps":["state f32","compute f32"],"group":118,"kernel":"causal_conv1d::causal_conv1d_update_l2norm_f32","family":"causal_conv1d_l2norm","point":null,"compute":"cuda_core","numerics":"reference","where":"device","state":"conv:update"},{"id":"l8.gdn.conv_ckpt","order":194,"layer":8,"block":"l8.gdn","block_type":"gdn","kind":"linear_attention","op":"state_snapshot","roles":[],"inserted":false,"precision":{"weight":null,"activation":"f32","output":null},"steps":["state f32"],"group":118,"kernel":"causal_conv1d::causal_conv1d_update_l2norm_f32","family":"causal_conv1d_l2norm","point":null,"compute":"cuda_core","numerics":"reference","where":"device","state":"conv:snapshot"},{"id":"l8.gdn.l2","order":195,"layer":8,"block":"l8.gdn","block_type":"gdn","kind":"linear_attention","op":"l2_norm","roles":[],"inserted":false,"precision":{"weight":null,"activation":"f32","output":"f32"},"steps":["compute f32"],"group":118,"kernel":"causal_conv1d::causal_conv1d_update_l2norm_f32","family":"causal_conv1d_l2norm","point":null,"compute":"cuda_core","numerics":"reference","where":"device","state":null},{"id":"l8.gdn.recur","order":196,"layer":8,"block":"l8.gdn","block_type":"gdn","kind":"linear_attention","op":"gdn_recurrence","roles":[],"inserted":false,"precision":{"weight":null,"activation":"f32","output":"f32"},"steps":["state f32","compute f32"],"group":119,"kernel":"gated_delta_rule::gated_delta_rule_decode_f32","family":"gdn_recurrence","point":{"decay":"head","gate_act":"silu"},"compute":"cuda_core","numerics":"reference","where":"device","state":"h:update"},{"id":"l8.gdn.out_norm","order":197,"layer":8,"block":"l8.gdn","block_type":"gdn","kind":"linear_attention","op":"gated_rms_norm","roles":[],"inserted":false,"precision":{"weight":null,"activation":"f32","output":"bf16"},"steps":["compute f32"],"group":120,"kernel":"norm::gated_rms_norm_f32_input","family":"gated_rms_norm","point":null,"compute":"memory","numerics":"reference","where":"device","state":null},{"id":"l8.gdn.out","order":198,"layer":8,"block":"l8.gdn","block_type":"gdn","kind":"linear_attention","op":"linear:gdn_out","roles":["gdn_out"],"inserted":false,"precision":{"weight":"fp8/block128x128","activation":"bf16","output":"bf16"},"steps":["act bf16","weight fp8/block128x128->bf16","mma bf16*bf16","accumulate f32","scale f32"],"group":121,"kernel":"w8a16_gemm_pipelined_m32::w8a16_gemm_pipelined_m32","family":"w8a16_gemm","point":null,"compute":"tensor_core:mma.sync.m16n8k16.bf16","numerics":"reference","where":"device","state":null},{"id":"l8.gdn.add","order":199,"layer":8,"block":"l8.gdn","block_type":"gdn","kind":"linear_attention","op":"residual_add","roles":[],"inserted":false,"precision":{"weight":null,"activation":"bf16","output":"bf16"},"steps":["compute f32"],"group":122,"kernel":"norm::residual_add_rms_norm","family":"rms_norm","point":null,"compute":"memory","numerics":"reference","where":"device","state":null},{"id":"l8.moe_ffn.post_norm","order":200,"layer":8,"block":"l8.moe_ffn","block_type":"moe_ffn","kind":"linear_attention","op":"rms_norm","roles":[],"inserted":false,"precision":{"weight":null,"activation":"f32","output":"bf16"},"steps":["compute f32"],"group":122,"kernel":"norm::residual_add_rms_norm","family":"rms_norm","point":null,"compute":"memory","numerics":"reference","where":"device","state":null},{"id":"l8.moe_ffn.router","order":201,"layer":8,"block":"l8.moe_ffn","block_type":"moe_ffn","kind":"linear_attention","op":"router","roles":[],"inserted":false,"precision":{"weight":"bf16","activation":"bf16","output":"bf16"},"steps":["act bf16","weight bf16->bf16","mma bf16*bf16","accumulate f32","scale none"],"group":123,"kernel":"gemv::dense_gemv_bf16","family":"dense_bf16","point":null,"compute":"cuda_core","numerics":"reference","where":"device","state":null},{"id":"l8.moe_ffn.top_k","order":202,"layer":8,"block":"l8.moe_ffn","block_type":"moe_ffn","kind":"linear_attention","op":"top_k","roles":[],"inserted":false,"precision":{"weight":null,"activation":"bf16","output":"f32,i32"},"steps":["score f32"],"group":124,"kernel":"moe_topk::moe_topk_softmax","family":"moe_topk","point":null,"compute":"memory","numerics":"reference","where":"device","state":null},{"id":"l8.moe_ffn.experts_gate_up","order":203,"layer":8,"block":"l8.moe_ffn","block_type":"moe_ffn","kind":"linear_attention","op":"expert_gate_up","roles":[],"inserted":false,"precision":{"weight":"fp8/block128x128","activation":"bf16","output":"bf16"},"steps":["gather bf16","act bf16","weight fp8/block128x128->bf16","mma bf16*bf16","accumulate f32","scale f32"],"group":125,"kernel":"moe_shared_expert_fused_fp8::moe_expert_gate_up_shared_fp8","family":"moe_fp8_1row","point":null,"compute":"cuda_core","numerics":"reference","where":"device","state":null},{"id":"l8.moe_ffn.shared_gate_up","order":204,"layer":8,"block":"l8.moe_ffn","block_type":"moe_ffn","kind":"linear_attention","op":"linear:shared_gate_up","roles":["shared_gate_up"],"inserted":false,"precision":{"weight":"fp8/block128x128","activation":"bf16","output":"bf16"},"steps":["act bf16","weight fp8/block128x128->bf16","mma bf16*bf16","accumulate f32","scale f32"],"group":125,"kernel":"moe_shared_expert_fused_fp8::moe_expert_gate_up_shared_fp8","family":"moe_fp8_1row","point":null,"compute":"cuda_core","numerics":"reference","where":"device","state":null},{"id":"l8.moe_ffn.experts_act","order":205,"layer":8,"block":"l8.moe_ffn","block_type":"moe_ffn","kind":"linear_attention","op":"silu_mul","roles":[],"inserted":false,"precision":{"weight":null,"activation":"bf16","output":"f32"},"steps":["compute f32"],"group":126,"kernel":"moe_shared_expert_fused_fp8::moe_expert_silu_down_shared_fp8","family":"moe_fp8_1row","point":null,"compute":"cuda_core","numerics":"reference","where":"device","state":null},{"id":"l8.moe_ffn.experts_down","order":206,"layer":8,"block":"l8.moe_ffn","block_type":"moe_ffn","kind":"linear_attention","op":"expert_down","roles":[],"inserted":false,"precision":{"weight":"fp8/block128x128","activation":"f32","output":"bf16"},"steps":["act f32","weight fp8/block128x128->f32","mma f32*f32","accumulate f32","scale f32"],"group":126,"kernel":"moe_shared_expert_fused_fp8::moe_expert_silu_down_shared_fp8","family":"moe_fp8_1row","point":null,"compute":"cuda_core","numerics":"reference","where":"device","state":null},{"id":"l8.moe_ffn.shared_act","order":207,"layer":8,"block":"l8.moe_ffn","block_type":"moe_ffn","kind":"linear_attention","op":"silu_mul","roles":[],"inserted":false,"precision":{"weight":null,"activation":"bf16","output":"f32"},"steps":["compute f32"],"group":126,"kernel":"moe_shared_expert_fused_fp8::moe_expert_silu_down_shared_fp8","family":"moe_fp8_1row","point":null,"compute":"cuda_core","numerics":"reference","where":"device","state":null},{"id":"l8.moe_ffn.shared_down","order":208,"layer":8,"block":"l8.moe_ffn","block_type":"moe_ffn","kind":"linear_attention","op":"linear:shared_down","roles":["shared_down"],"inserted":false,"precision":{"weight":"fp8/block128x128","activation":"f32","output":"bf16"},"steps":["act f32","weight fp8/block128x128->f32","mma f32*f32","accumulate f32","scale f32"],"group":126,"kernel":"moe_shared_expert_fused_fp8::moe_expert_silu_down_shared_fp8","family":"moe_fp8_1row","point":null,"compute":"cuda_core","numerics":"reference","where":"device","state":null},{"id":"l8.moe_ffn.shared_gate","order":209,"layer":8,"block":"l8.moe_ffn","block_type":"moe_ffn","kind":"linear_attention","op":"linear:shared_gate","roles":["shared_gate"],"inserted":false,"precision":{"weight":"bf16","activation":"bf16","output":"f32"},"steps":["act bf16","weight bf16->bf16","mma bf16*bf16","accumulate f32","scale none"],"group":127,"kernel":"moe_expert_gemv::moe_weighted_sum_blend","family":"moe_blend","point":{"shared_gate":"sigmoid"},"compute":"cuda_core","numerics":"reference","where":"device","state":null},{"id":"l8.moe_ffn.blend","order":210,"layer":8,"block":"l8.moe_ffn","block_type":"moe_ffn","kind":"linear_attention","op":"blend","roles":[],"inserted":false,"precision":{"weight":null,"activation":"bf16","output":"bf16"},"steps":["scatter f32","combine f32"],"group":127,"kernel":"moe_expert_gemv::moe_weighted_sum_blend","family":"moe_blend","point":{"shared_gate":"sigmoid"},"compute":"cuda_core","numerics":"reference","where":"device","state":null},{"id":"l8.moe_ffn.add","order":211,"layer":8,"block":"l8.moe_ffn","block_type":"moe_ffn","kind":"linear_attention","op":"residual_add","roles":[],"inserted":false,"precision":{"weight":null,"activation":"bf16","output":"bf16"},"steps":["compute f32"],"group":128,"kernel":"residual_add_rms_norm_exact::residual_add_rms_norm_exact","family":"rms_norm","point":null,"compute":"memory","numerics":"bit_identical(residual_add_rms_norm_exact_microtest)","where":"device","state":null},{"id":"l9.gdn.input_norm","order":212,"layer":9,"block":"l9.gdn","block_type":"gdn","kind":"linear_attention","op":"rms_norm","roles":[],"inserted":false,"precision":{"weight":null,"activation":"bf16","output":"bf16"},"steps":["compute f32"],"group":128,"kernel":"residual_add_rms_norm_exact::residual_add_rms_norm_exact","family":"rms_norm","point":null,"compute":"memory","numerics":"bit_identical(residual_add_rms_norm_exact_microtest)","where":"device","state":null},{"id":"l9.gdn.qkvz","order":213,"layer":9,"block":"l9.gdn","block_type":"gdn","kind":"linear_attention","op":"linear:qkvz","roles":["qkvz"],"inserted":false,"precision":{"weight":"fp8/block128x128","activation":"bf16","output":"bf16"},"steps":["act bf16","weight fp8/block128x128->bf16","mma bf16*bf16","accumulate f32","scale f32"],"group":129,"kernel":"w8a16_gemm_pipelined_m32::w8a16_gemm_pipelined_m32","family":"w8a16_gemm","point":null,"compute":"tensor_core:mma.sync.m16n8k16.bf16","numerics":"reference","where":"device","state":null},{"id":"l9.gdn.ba","order":214,"layer":9,"block":"l9.gdn","block_type":"gdn","kind":"linear_attention","op":"linear:ba","roles":["ba"],"inserted":false,"precision":{"weight":"bf16","activation":"bf16","output":"f32"},"steps":["act bf16","weight bf16->bf16","mma bf16*bf16","accumulate f32","scale none"],"group":130,"kernel":"ssm_preprocess::dense_gemv_ba_gates","family":"gdn_ba_gates_gemv","point":{"gate":"softplus","decay":"head"},"compute":"cuda_core","numerics":"reference","where":"device","state":null},{"id":"l9.gdn.gates","order":215,"layer":9,"block":"l9.gdn","block_type":"gdn","kind":"linear_attention","op":"gdn_gates","roles":[],"inserted":false,"precision":{"weight":null,"activation":"f32","output":"f32,f32"},"steps":["compute f32"],"group":130,"kernel":"ssm_preprocess::dense_gemv_ba_gates","family":"gdn_ba_gates_gemv","point":{"gate":"softplus","decay":"head"},"compute":"cuda_core","numerics":"reference","where":"device","state":null},{"id":"l9.gdn.conv","order":216,"layer":9,"block":"l9.gdn","block_type":"gdn","kind":"linear_attention","op":"conv1d_update","roles":[],"inserted":false,"precision":{"weight":null,"activation":"bf16","output":"f32"},"steps":["state f32","compute f32"],"group":131,"kernel":"causal_conv1d::causal_conv1d_update_l2norm_f32","family":"causal_conv1d_l2norm","point":null,"compute":"cuda_core","numerics":"reference","where":"device","state":"conv:update"},{"id":"l9.gdn.conv_ckpt","order":217,"layer":9,"block":"l9.gdn","block_type":"gdn","kind":"linear_attention","op":"state_snapshot","roles":[],"inserted":false,"precision":{"weight":null,"activation":"f32","output":null},"steps":["state f32"],"group":131,"kernel":"causal_conv1d::causal_conv1d_update_l2norm_f32","family":"causal_conv1d_l2norm","point":null,"compute":"cuda_core","numerics":"reference","where":"device","state":"conv:snapshot"},{"id":"l9.gdn.l2","order":218,"layer":9,"block":"l9.gdn","block_type":"gdn","kind":"linear_attention","op":"l2_norm","roles":[],"inserted":false,"precision":{"weight":null,"activation":"f32","output":"f32"},"steps":["compute f32"],"group":131,"kernel":"causal_conv1d::causal_conv1d_update_l2norm_f32","family":"causal_conv1d_l2norm","point":null,"compute":"cuda_core","numerics":"reference","where":"device","state":null},{"id":"l9.gdn.recur","order":219,"layer":9,"block":"l9.gdn","block_type":"gdn","kind":"linear_attention","op":"gdn_recurrence","roles":[],"inserted":false,"precision":{"weight":null,"activation":"f32","output":"f32"},"steps":["state f32","compute f32"],"group":132,"kernel":"gated_delta_rule::gated_delta_rule_decode_f32","family":"gdn_recurrence","point":{"decay":"head","gate_act":"silu"},"compute":"cuda_core","numerics":"reference","where":"device","state":"h:update"},{"id":"l9.gdn.out_norm","order":220,"layer":9,"block":"l9.gdn","block_type":"gdn","kind":"linear_attention","op":"gated_rms_norm","roles":[],"inserted":false,"precision":{"weight":null,"activation":"f32","output":"bf16"},"steps":["compute f32"],"group":133,"kernel":"norm::gated_rms_norm_f32_input","family":"gated_rms_norm","point":null,"compute":"memory","numerics":"reference","where":"device","state":null},{"id":"l9.gdn.out","order":221,"layer":9,"block":"l9.gdn","block_type":"gdn","kind":"linear_attention","op":"linear:gdn_out","roles":["gdn_out"],"inserted":false,"precision":{"weight":"fp8/block128x128","activation":"bf16","output":"bf16"},"steps":["act bf16","weight fp8/block128x128->bf16","mma bf16*bf16","accumulate f32","scale f32"],"group":134,"kernel":"w8a16_gemm_pipelined_m32::w8a16_gemm_pipelined_m32","family":"w8a16_gemm","point":null,"compute":"tensor_core:mma.sync.m16n8k16.bf16","numerics":"reference","where":"device","state":null},{"id":"l9.gdn.add","order":222,"layer":9,"block":"l9.gdn","block_type":"gdn","kind":"linear_attention","op":"residual_add","roles":[],"inserted":false,"precision":{"weight":null,"activation":"bf16","output":"bf16"},"steps":["compute f32"],"group":135,"kernel":"norm::residual_add_rms_norm","family":"rms_norm","point":null,"compute":"memory","numerics":"reference","where":"device","state":null},{"id":"l9.moe_ffn.post_norm","order":223,"layer":9,"block":"l9.moe_ffn","block_type":"moe_ffn","kind":"linear_attention","op":"rms_norm","roles":[],"inserted":false,"precision":{"weight":null,"activation":"f32","output":"bf16"},"steps":["compute f32"],"group":135,"kernel":"norm::residual_add_rms_norm","family":"rms_norm","point":null,"compute":"memory","numerics":"reference","where":"device","state":null},{"id":"l9.moe_ffn.router","order":224,"layer":9,"block":"l9.moe_ffn","block_type":"moe_ffn","kind":"linear_attention","op":"router","roles":[],"inserted":false,"precision":{"weight":"bf16","activation":"bf16","output":"bf16"},"steps":["act bf16","weight bf16->bf16","mma bf16*bf16","accumulate f32","scale none"],"group":136,"kernel":"gemv::dense_gemv_bf16","family":"dense_bf16","point":null,"compute":"cuda_core","numerics":"reference","where":"device","state":null},{"id":"l9.moe_ffn.top_k","order":225,"layer":9,"block":"l9.moe_ffn","block_type":"moe_ffn","kind":"linear_attention","op":"top_k","roles":[],"inserted":false,"precision":{"weight":null,"activation":"bf16","output":"f32,i32"},"steps":["score f32"],"group":137,"kernel":"moe_topk::moe_topk_softmax","family":"moe_topk","point":null,"compute":"memory","numerics":"reference","where":"device","state":null},{"id":"l9.moe_ffn.experts_gate_up","order":226,"layer":9,"block":"l9.moe_ffn","block_type":"moe_ffn","kind":"linear_attention","op":"expert_gate_up","roles":[],"inserted":false,"precision":{"weight":"fp8/block128x128","activation":"bf16","output":"bf16"},"steps":["gather bf16","act bf16","weight fp8/block128x128->bf16","mma bf16*bf16","accumulate f32","scale f32"],"group":138,"kernel":"moe_shared_expert_fused_fp8::moe_expert_gate_up_shared_fp8","family":"moe_fp8_1row","point":null,"compute":"cuda_core","numerics":"reference","where":"device","state":null},{"id":"l9.moe_ffn.shared_gate_up","order":227,"layer":9,"block":"l9.moe_ffn","block_type":"moe_ffn","kind":"linear_attention","op":"linear:shared_gate_up","roles":["shared_gate_up"],"inserted":false,"precision":{"weight":"fp8/block128x128","activation":"bf16","output":"bf16"},"steps":["act bf16","weight fp8/block128x128->bf16","mma bf16*bf16","accumulate f32","scale f32"],"group":138,"kernel":"moe_shared_expert_fused_fp8::moe_expert_gate_up_shared_fp8","family":"moe_fp8_1row","point":null,"compute":"cuda_core","numerics":"reference","where":"device","state":null},{"id":"l9.moe_ffn.experts_act","order":228,"layer":9,"block":"l9.moe_ffn","block_type":"moe_ffn","kind":"linear_attention","op":"silu_mul","roles":[],"inserted":false,"precision":{"weight":null,"activation":"bf16","output":"f32"},"steps":["compute f32"],"group":139,"kernel":"moe_shared_expert_fused_fp8::moe_expert_silu_down_shared_fp8","family":"moe_fp8_1row","point":null,"compute":"cuda_core","numerics":"reference","where":"device","state":null},{"id":"l9.moe_ffn.experts_down","order":229,"layer":9,"block":"l9.moe_ffn","block_type":"moe_ffn","kind":"linear_attention","op":"expert_down","roles":[],"inserted":false,"precision":{"weight":"fp8/block128x128","activation":"f32","output":"bf16"},"steps":["act f32","weight fp8/block128x128->f32","mma f32*f32","accumulate f32","scale f32"],"group":139,"kernel":"moe_shared_expert_fused_fp8::moe_expert_silu_down_shared_fp8","family":"moe_fp8_1row","point":null,"compute":"cuda_core","numerics":"reference","where":"device","state":null},{"id":"l9.moe_ffn.shared_act","order":230,"layer":9,"block":"l9.moe_ffn","block_type":"moe_ffn","kind":"linear_attention","op":"silu_mul","roles":[],"inserted":false,"precision":{"weight":null,"activation":"bf16","output":"f32"},"steps":["compute f32"],"group":139,"kernel":"moe_shared_expert_fused_fp8::moe_expert_silu_down_shared_fp8","family":"moe_fp8_1row","point":null,"compute":"cuda_core","numerics":"reference","where":"device","state":null},{"id":"l9.moe_ffn.shared_down","order":231,"layer":9,"block":"l9.moe_ffn","block_type":"moe_ffn","kind":"linear_attention","op":"linear:shared_down","roles":["shared_down"],"inserted":false,"precision":{"weight":"fp8/block128x128","activation":"f32","output":"bf16"},"steps":["act f32","weight fp8/block128x128->f32","mma f32*f32","accumulate f32","scale f32"],"group":139,"kernel":"moe_shared_expert_fused_fp8::moe_expert_silu_down_shared_fp8","family":"moe_fp8_1row","point":null,"compute":"cuda_core","numerics":"reference","where":"device","state":null},{"id":"l9.moe_ffn.shared_gate","order":232,"layer":9,"block":"l9.moe_ffn","block_type":"moe_ffn","kind":"linear_attention","op":"linear:shared_gate","roles":["shared_gate"],"inserted":false,"precision":{"weight":"bf16","activation":"bf16","output":"f32"},"steps":["act bf16","weight bf16->bf16","mma bf16*bf16","accumulate f32","scale none"],"group":140,"kernel":"moe_expert_gemv::moe_weighted_sum_blend","family":"moe_blend","point":{"shared_gate":"sigmoid"},"compute":"cuda_core","numerics":"reference","where":"device","state":null},{"id":"l9.moe_ffn.blend","order":233,"layer":9,"block":"l9.moe_ffn","block_type":"moe_ffn","kind":"linear_attention","op":"blend","roles":[],"inserted":false,"precision":{"weight":null,"activation":"bf16","output":"bf16"},"steps":["scatter f32","combine f32"],"group":140,"kernel":"moe_expert_gemv::moe_weighted_sum_blend","family":"moe_blend","point":{"shared_gate":"sigmoid"},"compute":"cuda_core","numerics":"reference","where":"device","state":null},{"id":"l9.moe_ffn.add","order":234,"layer":9,"block":"l9.moe_ffn","block_type":"moe_ffn","kind":"linear_attention","op":"residual_add","roles":[],"inserted":false,"precision":{"weight":null,"activation":"bf16","output":"bf16"},"steps":["compute f32"],"group":141,"kernel":"residual_add_rms_norm_exact::residual_add_rms_norm_exact","family":"rms_norm","point":null,"compute":"memory","numerics":"bit_identical(residual_add_rms_norm_exact_microtest)","where":"device","state":null},{"id":"l10.gdn.input_norm","order":235,"layer":10,"block":"l10.gdn","block_type":"gdn","kind":"linear_attention","op":"rms_norm","roles":[],"inserted":false,"precision":{"weight":null,"activation":"bf16","output":"bf16"},"steps":["compute f32"],"group":141,"kernel":"residual_add_rms_norm_exact::residual_add_rms_norm_exact","family":"rms_norm","point":null,"compute":"memory","numerics":"bit_identical(residual_add_rms_norm_exact_microtest)","where":"device","state":null},{"id":"l10.gdn.qkvz","order":236,"layer":10,"block":"l10.gdn","block_type":"gdn","kind":"linear_attention","op":"linear:qkvz","roles":["qkvz"],"inserted":false,"precision":{"weight":"fp8/block128x128","activation":"bf16","output":"bf16"},"steps":["act bf16","weight fp8/block128x128->bf16","mma bf16*bf16","accumulate f32","scale f32"],"group":142,"kernel":"w8a16_gemm_pipelined_m32::w8a16_gemm_pipelined_m32","family":"w8a16_gemm","point":null,"compute":"tensor_core:mma.sync.m16n8k16.bf16","numerics":"reference","where":"device","state":null},{"id":"l10.gdn.ba","order":237,"layer":10,"block":"l10.gdn","block_type":"gdn","kind":"linear_attention","op":"linear:ba","roles":["ba"],"inserted":false,"precision":{"weight":"bf16","activation":"bf16","output":"f32"},"steps":["act bf16","weight bf16->bf16","mma bf16*bf16","accumulate f32","scale none"],"group":143,"kernel":"ssm_preprocess::dense_gemv_ba_gates","family":"gdn_ba_gates_gemv","point":{"gate":"softplus","decay":"head"},"compute":"cuda_core","numerics":"reference","where":"device","state":null},{"id":"l10.gdn.gates","order":238,"layer":10,"block":"l10.gdn","block_type":"gdn","kind":"linear_attention","op":"gdn_gates","roles":[],"inserted":false,"precision":{"weight":null,"activation":"f32","output":"f32,f32"},"steps":["compute f32"],"group":143,"kernel":"ssm_preprocess::dense_gemv_ba_gates","family":"gdn_ba_gates_gemv","point":{"gate":"softplus","decay":"head"},"compute":"cuda_core","numerics":"reference","where":"device","state":null},{"id":"l10.gdn.conv","order":239,"layer":10,"block":"l10.gdn","block_type":"gdn","kind":"linear_attention","op":"conv1d_update","roles":[],"inserted":false,"precision":{"weight":null,"activation":"bf16","output":"f32"},"steps":["state f32","compute f32"],"group":144,"kernel":"causal_conv1d::causal_conv1d_update_l2norm_f32","family":"causal_conv1d_l2norm","point":null,"compute":"cuda_core","numerics":"reference","where":"device","state":"conv:update"},{"id":"l10.gdn.conv_ckpt","order":240,"layer":10,"block":"l10.gdn","block_type":"gdn","kind":"linear_attention","op":"state_snapshot","roles":[],"inserted":false,"precision":{"weight":null,"activation":"f32","output":null},"steps":["state f32"],"group":144,"kernel":"causal_conv1d::causal_conv1d_update_l2norm_f32","family":"causal_conv1d_l2norm","point":null,"compute":"cuda_core","numerics":"reference","where":"device","state":"conv:snapshot"},{"id":"l10.gdn.l2","order":241,"layer":10,"block":"l10.gdn","block_type":"gdn","kind":"linear_attention","op":"l2_norm","roles":[],"inserted":false,"precision":{"weight":null,"activation":"f32","output":"f32"},"steps":["compute f32"],"group":144,"kernel":"causal_conv1d::causal_conv1d_update_l2norm_f32","family":"causal_conv1d_l2norm","point":null,"compute":"cuda_core","numerics":"reference","where":"device","state":null},{"id":"l10.gdn.recur","order":242,"layer":10,"block":"l10.gdn","block_type":"gdn","kind":"linear_attention","op":"gdn_recurrence","roles":[],"inserted":false,"precision":{"weight":null,"activation":"f32","output":"f32"},"steps":["state f32","compute f32"],"group":145,"kernel":"gated_delta_rule::gated_delta_rule_decode_f32","family":"gdn_recurrence","point":{"decay":"head","gate_act":"silu"},"compute":"cuda_core","numerics":"reference","where":"device","state":"h:update"},{"id":"l10.gdn.out_norm","order":243,"layer":10,"block":"l10.gdn","block_type":"gdn","kind":"linear_attention","op":"gated_rms_norm","roles":[],"inserted":false,"precision":{"weight":null,"activation":"f32","output":"bf16"},"steps":["compute f32"],"group":146,"kernel":"norm::gated_rms_norm_f32_input","family":"gated_rms_norm","point":null,"compute":"memory","numerics":"reference","where":"device","state":null},{"id":"l10.gdn.out","order":244,"layer":10,"block":"l10.gdn","block_type":"gdn","kind":"linear_attention","op":"linear:gdn_out","roles":["gdn_out"],"inserted":false,"precision":{"weight":"fp8/block128x128","activation":"bf16","output":"bf16"},"steps":["act bf16","weight fp8/block128x128->bf16","mma bf16*bf16","accumulate f32","scale f32"],"group":147,"kernel":"w8a16_gemm_pipelined_m32::w8a16_gemm_pipelined_m32","family":"w8a16_gemm","point":null,"compute":"tensor_core:mma.sync.m16n8k16.bf16","numerics":"reference","where":"device","state":null},{"id":"l10.gdn.add","order":245,"layer":10,"block":"l10.gdn","block_type":"gdn","kind":"linear_attention","op":"residual_add","roles":[],"inserted":false,"precision":{"weight":null,"activation":"bf16","output":"bf16"},"steps":["compute f32"],"group":148,"kernel":"norm::residual_add_rms_norm","family":"rms_norm","point":null,"compute":"memory","numerics":"reference","where":"device","state":null},{"id":"l10.moe_ffn.post_norm","order":246,"layer":10,"block":"l10.moe_ffn","block_type":"moe_ffn","kind":"linear_attention","op":"rms_norm","roles":[],"inserted":false,"precision":{"weight":null,"activation":"f32","output":"bf16"},"steps":["compute f32"],"group":148,"kernel":"norm::residual_add_rms_norm","family":"rms_norm","point":null,"compute":"memory","numerics":"reference","where":"device","state":null},{"id":"l10.moe_ffn.router","order":247,"layer":10,"block":"l10.moe_ffn","block_type":"moe_ffn","kind":"linear_attention","op":"router","roles":[],"inserted":false,"precision":{"weight":"bf16","activation":"bf16","output":"bf16"},"steps":["act bf16","weight bf16->bf16","mma bf16*bf16","accumulate f32","scale none"],"group":149,"kernel":"gemv::dense_gemv_bf16","family":"dense_bf16","point":null,"compute":"cuda_core","numerics":"reference","where":"device","state":null},{"id":"l10.moe_ffn.top_k","order":248,"layer":10,"block":"l10.moe_ffn","block_type":"moe_ffn","kind":"linear_attention","op":"top_k","roles":[],"inserted":false,"precision":{"weight":null,"activation":"bf16","output":"f32,i32"},"steps":["score f32"],"group":150,"kernel":"moe_topk::moe_topk_softmax","family":"moe_topk","point":null,"compute":"memory","numerics":"reference","where":"device","state":null},{"id":"l10.moe_ffn.experts_gate_up","order":249,"layer":10,"block":"l10.moe_ffn","block_type":"moe_ffn","kind":"linear_attention","op":"expert_gate_up","roles":[],"inserted":false,"precision":{"weight":"fp8/block128x128","activation":"bf16","output":"bf16"},"steps":["gather bf16","act bf16","weight fp8/block128x128->bf16","mma bf16*bf16","accumulate f32","scale f32"],"group":151,"kernel":"moe_shared_expert_fused_fp8::moe_expert_gate_up_shared_fp8","family":"moe_fp8_1row","point":null,"compute":"cuda_core","numerics":"reference","where":"device","state":null},{"id":"l10.moe_ffn.shared_gate_up","order":250,"layer":10,"block":"l10.moe_ffn","block_type":"moe_ffn","kind":"linear_attention","op":"linear:shared_gate_up","roles":["shared_gate_up"],"inserted":false,"precision":{"weight":"fp8/block128x128","activation":"bf16","output":"bf16"},"steps":["act bf16","weight fp8/block128x128->bf16","mma bf16*bf16","accumulate f32","scale f32"],"group":151,"kernel":"moe_shared_expert_fused_fp8::moe_expert_gate_up_shared_fp8","family":"moe_fp8_1row","point":null,"compute":"cuda_core","numerics":"reference","where":"device","state":null},{"id":"l10.moe_ffn.experts_act","order":251,"layer":10,"block":"l10.moe_ffn","block_type":"moe_ffn","kind":"linear_attention","op":"silu_mul","roles":[],"inserted":false,"precision":{"weight":null,"activation":"bf16","output":"f32"},"steps":["compute f32"],"group":152,"kernel":"moe_shared_expert_fused_fp8::moe_expert_silu_down_shared_fp8","family":"moe_fp8_1row","point":null,"compute":"cuda_core","numerics":"reference","where":"device","state":null},{"id":"l10.moe_ffn.experts_down","order":252,"layer":10,"block":"l10.moe_ffn","block_type":"moe_ffn","kind":"linear_attention","op":"expert_down","roles":[],"inserted":false,"precision":{"weight":"fp8/block128x128","activation":"f32","output":"bf16"},"steps":["act f32","weight fp8/block128x128->f32","mma f32*f32","accumulate f32","scale f32"],"group":152,"kernel":"moe_shared_expert_fused_fp8::moe_expert_silu_down_shared_fp8","family":"moe_fp8_1row","point":null,"compute":"cuda_core","numerics":"reference","where":"device","state":null},{"id":"l10.moe_ffn.shared_act","order":253,"layer":10,"block":"l10.moe_ffn","block_type":"moe_ffn","kind":"linear_attention","op":"silu_mul","roles":[],"inserted":false,"precision":{"weight":null,"activation":"bf16","output":"f32"},"steps":["compute f32"],"group":152,"kernel":"moe_shared_expert_fused_fp8::moe_expert_silu_down_shared_fp8","family":"moe_fp8_1row","point":null,"compute":"cuda_core","numerics":"reference","where":"device","state":null},{"id":"l10.moe_ffn.shared_down","order":254,"layer":10,"block":"l10.moe_ffn","block_type":"moe_ffn","kind":"linear_attention","op":"linear:shared_down","roles":["shared_down"],"inserted":false,"precision":{"weight":"fp8/block128x128","activation":"f32","output":"bf16"},"steps":["act f32","weight fp8/block128x128->f32","mma f32*f32","accumulate f32","scale f32"],"group":152,"kernel":"moe_shared_expert_fused_fp8::moe_expert_silu_down_shared_fp8","family":"moe_fp8_1row","point":null,"compute":"cuda_core","numerics":"reference","where":"device","state":null},{"id":"l10.moe_ffn.shared_gate","order":255,"layer":10,"block":"l10.moe_ffn","block_type":"moe_ffn","kind":"linear_attention","op":"linear:shared_gate","roles":["shared_gate"],"inserted":false,"precision":{"weight":"bf16","activation":"bf16","output":"f32"},"steps":["act bf16","weight bf16->bf16","mma bf16*bf16","accumulate f32","scale none"],"group":153,"kernel":"moe_expert_gemv::moe_weighted_sum_blend","family":"moe_blend","point":{"shared_gate":"sigmoid"},"compute":"cuda_core","numerics":"reference","where":"device","state":null},{"id":"l10.moe_ffn.blend","order":256,"layer":10,"block":"l10.moe_ffn","block_type":"moe_ffn","kind":"linear_attention","op":"blend","roles":[],"inserted":false,"precision":{"weight":null,"activation":"bf16","output":"bf16"},"steps":["scatter f32","combine f32"],"group":153,"kernel":"moe_expert_gemv::moe_weighted_sum_blend","family":"moe_blend","point":{"shared_gate":"sigmoid"},"compute":"cuda_core","numerics":"reference","where":"device","state":null},{"id":"l10.moe_ffn.add","order":257,"layer":10,"block":"l10.moe_ffn","block_type":"moe_ffn","kind":"linear_attention","op":"residual_add","roles":[],"inserted":false,"precision":{"weight":null,"activation":"bf16","output":"bf16"},"steps":["compute f32"],"group":154,"kernel":"residual_add_rms_norm_exact::residual_add_rms_norm_exact","family":"rms_norm","point":null,"compute":"memory","numerics":"bit_identical(residual_add_rms_norm_exact_microtest)","where":"device","state":null},{"id":"l11.attn.input_norm","order":258,"layer":11,"block":"l11.attn","block_type":"attn","kind":"full_attention","op":"rms_norm","roles":[],"inserted":false,"precision":{"weight":null,"activation":"bf16","output":"bf16"},"steps":["compute f32"],"group":154,"kernel":"residual_add_rms_norm_exact::residual_add_rms_norm_exact","family":"rms_norm","point":null,"compute":"memory","numerics":"bit_identical(residual_add_rms_norm_exact_microtest)","where":"device","state":null},{"id":"l11.attn.q","order":259,"layer":11,"block":"l11.attn","block_type":"attn","kind":"full_attention","op":"linear:q","roles":["q"],"inserted":false,"precision":{"weight":"fp8/block128x128","activation":"bf16","output":"bf16"},"steps":["act bf16","weight fp8/block128x128->bf16","mma bf16*bf16","accumulate f32","scale f32"],"group":155,"kernel":"w8a16_gemm_pipelined_m32::w8a16_gemm_pipelined_m32","family":"w8a16_gemm","point":null,"compute":"tensor_core:mma.sync.m16n8k16.bf16","numerics":"reference","where":"device","state":null},{"id":"l11.attn.q_split","order":260,"layer":11,"block":"l11.attn","block_type":"attn","kind":"full_attention","op":"split","roles":[],"inserted":false,"precision":{"weight":null,"activation":"bf16","output":"bf16,bf16"},"steps":["move bf16"],"group":156,"kernel":"ssm_preprocess::deinterleave_qg","family":"deinterleave_qg","point":{"layout":"interleaved"},"compute":"memory","numerics":"reference","where":"device","state":null},{"id":"l11.attn.k","order":261,"layer":11,"block":"l11.attn","block_type":"attn","kind":"full_attention","op":"linear:k","roles":["k"],"inserted":false,"precision":{"weight":"fp8/block128x128","activation":"bf16","output":"bf16"},"steps":["act bf16","weight fp8/block128x128->bf16","mma bf16*bf16","accumulate f32","scale f32"],"group":157,"kernel":"w8a16_gemm_pipelined_m32::w8a16_gemm_pipelined_m32","family":"w8a16_gemm","point":null,"compute":"tensor_core:mma.sync.m16n8k16.bf16","numerics":"reference","where":"device","state":null},{"id":"l11.attn.v","order":262,"layer":11,"block":"l11.attn","block_type":"attn","kind":"full_attention","op":"linear:v","roles":["v"],"inserted":false,"precision":{"weight":"fp8/block128x128","activation":"bf16","output":"bf16"},"steps":["act bf16","weight fp8/block128x128->bf16","mma bf16*bf16","accumulate f32","scale f32"],"group":158,"kernel":"w8a16_gemm_pipelined_m32::w8a16_gemm_pipelined_m32","family":"w8a16_gemm","point":null,"compute":"tensor_core:mma.sync.m16n8k16.bf16","numerics":"reference","where":"device","state":null},{"id":"l11.attn.q_norm","order":263,"layer":11,"block":"l11.attn","block_type":"attn","kind":"full_attention","op":"qk_norm","roles":[],"inserted":false,"precision":{"weight":null,"activation":"bf16","output":"bf16"},"steps":["compute f32"],"group":159,"kernel":"norm::rms_norm","family":"rms_norm","point":null,"compute":"memory","numerics":"reference","where":"device","state":null},{"id":"l11.attn.k_norm","order":264,"layer":11,"block":"l11.attn","block_type":"attn","kind":"full_attention","op":"qk_norm","roles":[],"inserted":false,"precision":{"weight":null,"activation":"bf16","output":"bf16"},"steps":["compute f32"],"group":160,"kernel":"norm::rms_norm","family":"rms_norm","point":null,"compute":"memory","numerics":"reference","where":"device","state":null},{"id":"l11.attn.rope","order":265,"layer":11,"block":"l11.attn","block_type":"attn","kind":"full_attention","op":"rope","roles":[],"inserted":false,"precision":{"weight":null,"activation":"bf16","output":"bf16,bf16"},"steps":["compute f32"],"group":161,"kernel":"rope_mrope_interleaved::rope_forward_mrope_interleaved","family":"rope","point":null,"compute":"memory","numerics":"reference","where":"device","state":null},{"id":"l11.attn.kv_write","order":266,"layer":11,"block":"l11.attn","block_type":"attn","kind":"full_attention","op":"kv_write","roles":[],"inserted":false,"precision":{"weight":null,"activation":"bf16","output":null},"steps":["cache bf16"],"group":162,"kernel":"reshape_and_cache::reshape_and_cache_flash","family":"kv_write","point":null,"compute":"memory","numerics":"reference","where":"device","state":"k:write,v:write"},{"id":"l11.attn.attend","order":267,"layer":11,"block":"l11.attn","block_type":"attn","kind":"full_attention","op":"paged_attention","roles":[],"inserted":false,"precision":{"weight":null,"activation":"bf16","output":"bf16"},"steps":["cache bf16","scores f32","softmax f32","accumulate f32"],"group":163,"kernel":"paged_decode::paged_decode_attn","family":"paged_decode_attn","point":null,"compute":"cuda_core","numerics":"reference","where":"device","state":"k:read,v:read"},{"id":"l11.attn.gate_mul","order":268,"layer":11,"block":"l11.attn","block_type":"attn","kind":"full_attention","op":"sigmoid_gate_mul","roles":[],"inserted":false,"precision":{"weight":null,"activation":"bf16","output":"bf16"},"steps":["compute f32"],"group":164,"kernel":"residual_add::sigmoid_gate_mul","family":"sigmoid_gate_mul","point":null,"compute":"memory","numerics":"reference","where":"device","state":null},{"id":"l11.attn.o","order":269,"layer":11,"block":"l11.attn","block_type":"attn","kind":"full_attention","op":"linear:o","roles":["o"],"inserted":false,"precision":{"weight":"fp8/block128x128","activation":"bf16","output":"bf16"},"steps":["act bf16","weight fp8/block128x128->bf16","mma bf16*bf16","accumulate f32","scale f32"],"group":165,"kernel":"w8a16_gemm_pipelined_m32::w8a16_gemm_pipelined_m32","family":"w8a16_gemm","point":null,"compute":"tensor_core:mma.sync.m16n8k16.bf16","numerics":"reference","where":"device","state":null},{"id":"l11.attn.add","order":270,"layer":11,"block":"l11.attn","block_type":"attn","kind":"full_attention","op":"residual_add","roles":[],"inserted":false,"precision":{"weight":null,"activation":"bf16","output":"bf16"},"steps":["compute f32"],"group":166,"kernel":"norm::residual_add_rms_norm","family":"rms_norm","point":null,"compute":"memory","numerics":"reference","where":"device","state":null},{"id":"l11.moe_ffn.post_norm","order":271,"layer":11,"block":"l11.moe_ffn","block_type":"moe_ffn","kind":"full_attention","op":"rms_norm","roles":[],"inserted":false,"precision":{"weight":null,"activation":"f32","output":"bf16"},"steps":["compute f32"],"group":166,"kernel":"norm::residual_add_rms_norm","family":"rms_norm","point":null,"compute":"memory","numerics":"reference","where":"device","state":null},{"id":"l11.moe_ffn.router","order":272,"layer":11,"block":"l11.moe_ffn","block_type":"moe_ffn","kind":"full_attention","op":"router","roles":[],"inserted":false,"precision":{"weight":"bf16","activation":"bf16","output":"bf16"},"steps":["act bf16","weight bf16->bf16","mma bf16*bf16","accumulate f32","scale none"],"group":167,"kernel":"gemv::dense_gemv_bf16","family":"dense_bf16","point":null,"compute":"cuda_core","numerics":"reference","where":"device","state":null},{"id":"l11.moe_ffn.top_k","order":273,"layer":11,"block":"l11.moe_ffn","block_type":"moe_ffn","kind":"full_attention","op":"top_k","roles":[],"inserted":false,"precision":{"weight":null,"activation":"bf16","output":"f32,i32"},"steps":["score f32"],"group":168,"kernel":"moe_topk::moe_topk_softmax","family":"moe_topk","point":null,"compute":"memory","numerics":"reference","where":"device","state":null},{"id":"l11.moe_ffn.experts_gate_up","order":274,"layer":11,"block":"l11.moe_ffn","block_type":"moe_ffn","kind":"full_attention","op":"expert_gate_up","roles":[],"inserted":false,"precision":{"weight":"fp8/block128x128","activation":"bf16","output":"bf16"},"steps":["gather bf16","act bf16","weight fp8/block128x128->bf16","mma bf16*bf16","accumulate f32","scale f32"],"group":169,"kernel":"moe_shared_expert_fused_fp8::moe_expert_gate_up_shared_fp8","family":"moe_fp8_1row","point":null,"compute":"cuda_core","numerics":"reference","where":"device","state":null},{"id":"l11.moe_ffn.shared_gate_up","order":275,"layer":11,"block":"l11.moe_ffn","block_type":"moe_ffn","kind":"full_attention","op":"linear:shared_gate_up","roles":["shared_gate_up"],"inserted":false,"precision":{"weight":"fp8/block128x128","activation":"bf16","output":"bf16"},"steps":["act bf16","weight fp8/block128x128->bf16","mma bf16*bf16","accumulate f32","scale f32"],"group":169,"kernel":"moe_shared_expert_fused_fp8::moe_expert_gate_up_shared_fp8","family":"moe_fp8_1row","point":null,"compute":"cuda_core","numerics":"reference","where":"device","state":null},{"id":"l11.moe_ffn.experts_act","order":276,"layer":11,"block":"l11.moe_ffn","block_type":"moe_ffn","kind":"full_attention","op":"silu_mul","roles":[],"inserted":false,"precision":{"weight":null,"activation":"bf16","output":"f32"},"steps":["compute f32"],"group":170,"kernel":"moe_shared_expert_fused_fp8::moe_expert_silu_down_shared_fp8","family":"moe_fp8_1row","point":null,"compute":"cuda_core","numerics":"reference","where":"device","state":null},{"id":"l11.moe_ffn.experts_down","order":277,"layer":11,"block":"l11.moe_ffn","block_type":"moe_ffn","kind":"full_attention","op":"expert_down","roles":[],"inserted":false,"precision":{"weight":"fp8/block128x128","activation":"f32","output":"bf16"},"steps":["act f32","weight fp8/block128x128->f32","mma f32*f32","accumulate f32","scale f32"],"group":170,"kernel":"moe_shared_expert_fused_fp8::moe_expert_silu_down_shared_fp8","family":"moe_fp8_1row","point":null,"compute":"cuda_core","numerics":"reference","where":"device","state":null},{"id":"l11.moe_ffn.shared_act","order":278,"layer":11,"block":"l11.moe_ffn","block_type":"moe_ffn","kind":"full_attention","op":"silu_mul","roles":[],"inserted":false,"precision":{"weight":null,"activation":"bf16","output":"f32"},"steps":["compute f32"],"group":170,"kernel":"moe_shared_expert_fused_fp8::moe_expert_silu_down_shared_fp8","family":"moe_fp8_1row","point":null,"compute":"cuda_core","numerics":"reference","where":"device","state":null},{"id":"l11.moe_ffn.shared_down","order":279,"layer":11,"block":"l11.moe_ffn","block_type":"moe_ffn","kind":"full_attention","op":"linear:shared_down","roles":["shared_down"],"inserted":false,"precision":{"weight":"fp8/block128x128","activation":"f32","output":"bf16"},"steps":["act f32","weight fp8/block128x128->f32","mma f32*f32","accumulate f32","scale f32"],"group":170,"kernel":"moe_shared_expert_fused_fp8::moe_expert_silu_down_shared_fp8","family":"moe_fp8_1row","point":null,"compute":"cuda_core","numerics":"reference","where":"device","state":null},{"id":"l11.moe_ffn.shared_gate","order":280,"layer":11,"block":"l11.moe_ffn","block_type":"moe_ffn","kind":"full_attention","op":"linear:shared_gate","roles":["shared_gate"],"inserted":false,"precision":{"weight":"bf16","activation":"bf16","output":"f32"},"steps":["act bf16","weight bf16->bf16","mma bf16*bf16","accumulate f32","scale none"],"group":171,"kernel":"moe_expert_gemv::moe_weighted_sum_blend","family":"moe_blend","point":{"shared_gate":"sigmoid"},"compute":"cuda_core","numerics":"reference","where":"device","state":null},{"id":"l11.moe_ffn.blend","order":281,"layer":11,"block":"l11.moe_ffn","block_type":"moe_ffn","kind":"full_attention","op":"blend","roles":[],"inserted":false,"precision":{"weight":null,"activation":"bf16","output":"bf16"},"steps":["scatter f32","combine f32"],"group":171,"kernel":"moe_expert_gemv::moe_weighted_sum_blend","family":"moe_blend","point":{"shared_gate":"sigmoid"},"compute":"cuda_core","numerics":"reference","where":"device","state":null},{"id":"l11.moe_ffn.add","order":282,"layer":11,"block":"l11.moe_ffn","block_type":"moe_ffn","kind":"full_attention","op":"residual_add","roles":[],"inserted":false,"precision":{"weight":null,"activation":"bf16","output":"bf16"},"steps":["compute f32"],"group":172,"kernel":"residual_add_rms_norm_exact::residual_add_rms_norm_exact","family":"rms_norm","point":null,"compute":"memory","numerics":"bit_identical(residual_add_rms_norm_exact_microtest)","where":"device","state":null},{"id":"l12.gdn.input_norm","order":283,"layer":12,"block":"l12.gdn","block_type":"gdn","kind":"linear_attention","op":"rms_norm","roles":[],"inserted":false,"precision":{"weight":null,"activation":"bf16","output":"bf16"},"steps":["compute f32"],"group":172,"kernel":"residual_add_rms_norm_exact::residual_add_rms_norm_exact","family":"rms_norm","point":null,"compute":"memory","numerics":"bit_identical(residual_add_rms_norm_exact_microtest)","where":"device","state":null},{"id":"l12.gdn.qkvz","order":284,"layer":12,"block":"l12.gdn","block_type":"gdn","kind":"linear_attention","op":"linear:qkvz","roles":["qkvz"],"inserted":false,"precision":{"weight":"fp8/block128x128","activation":"bf16","output":"bf16"},"steps":["act bf16","weight fp8/block128x128->bf16","mma bf16*bf16","accumulate f32","scale f32"],"group":173,"kernel":"w8a16_gemm_pipelined_m32::w8a16_gemm_pipelined_m32","family":"w8a16_gemm","point":null,"compute":"tensor_core:mma.sync.m16n8k16.bf16","numerics":"reference","where":"device","state":null},{"id":"l12.gdn.ba","order":285,"layer":12,"block":"l12.gdn","block_type":"gdn","kind":"linear_attention","op":"linear:ba","roles":["ba"],"inserted":false,"precision":{"weight":"bf16","activation":"bf16","output":"f32"},"steps":["act bf16","weight bf16->bf16","mma bf16*bf16","accumulate f32","scale none"],"group":174,"kernel":"ssm_preprocess::dense_gemv_ba_gates","family":"gdn_ba_gates_gemv","point":{"gate":"softplus","decay":"head"},"compute":"cuda_core","numerics":"reference","where":"device","state":null},{"id":"l12.gdn.gates","order":286,"layer":12,"block":"l12.gdn","block_type":"gdn","kind":"linear_attention","op":"gdn_gates","roles":[],"inserted":false,"precision":{"weight":null,"activation":"f32","output":"f32,f32"},"steps":["compute f32"],"group":174,"kernel":"ssm_preprocess::dense_gemv_ba_gates","family":"gdn_ba_gates_gemv","point":{"gate":"softplus","decay":"head"},"compute":"cuda_core","numerics":"reference","where":"device","state":null},{"id":"l12.gdn.conv","order":287,"layer":12,"block":"l12.gdn","block_type":"gdn","kind":"linear_attention","op":"conv1d_update","roles":[],"inserted":false,"precision":{"weight":null,"activation":"bf16","output":"f32"},"steps":["state f32","compute f32"],"group":175,"kernel":"causal_conv1d::causal_conv1d_update_l2norm_f32","family":"causal_conv1d_l2norm","point":null,"compute":"cuda_core","numerics":"reference","where":"device","state":"conv:update"},{"id":"l12.gdn.conv_ckpt","order":288,"layer":12,"block":"l12.gdn","block_type":"gdn","kind":"linear_attention","op":"state_snapshot","roles":[],"inserted":false,"precision":{"weight":null,"activation":"f32","output":null},"steps":["state f32"],"group":175,"kernel":"causal_conv1d::causal_conv1d_update_l2norm_f32","family":"causal_conv1d_l2norm","point":null,"compute":"cuda_core","numerics":"reference","where":"device","state":"conv:snapshot"},{"id":"l12.gdn.l2","order":289,"layer":12,"block":"l12.gdn","block_type":"gdn","kind":"linear_attention","op":"l2_norm","roles":[],"inserted":false,"precision":{"weight":null,"activation":"f32","output":"f32"},"steps":["compute f32"],"group":175,"kernel":"causal_conv1d::causal_conv1d_update_l2norm_f32","family":"causal_conv1d_l2norm","point":null,"compute":"cuda_core","numerics":"reference","where":"device","state":null},{"id":"l12.gdn.recur","order":290,"layer":12,"block":"l12.gdn","block_type":"gdn","kind":"linear_attention","op":"gdn_recurrence","roles":[],"inserted":false,"precision":{"weight":null,"activation":"f32","output":"f32"},"steps":["state f32","compute f32"],"group":176,"kernel":"gated_delta_rule::gated_delta_rule_decode_f32","family":"gdn_recurrence","point":{"decay":"head","gate_act":"silu"},"compute":"cuda_core","numerics":"reference","where":"device","state":"h:update"},{"id":"l12.gdn.out_norm","order":291,"layer":12,"block":"l12.gdn","block_type":"gdn","kind":"linear_attention","op":"gated_rms_norm","roles":[],"inserted":false,"precision":{"weight":null,"activation":"f32","output":"bf16"},"steps":["compute f32"],"group":177,"kernel":"norm::gated_rms_norm_f32_input","family":"gated_rms_norm","point":null,"compute":"memory","numerics":"reference","where":"device","state":null},{"id":"l12.gdn.out","order":292,"layer":12,"block":"l12.gdn","block_type":"gdn","kind":"linear_attention","op":"linear:gdn_out","roles":["gdn_out"],"inserted":false,"precision":{"weight":"fp8/block128x128","activation":"bf16","output":"bf16"},"steps":["act bf16","weight fp8/block128x128->bf16","mma bf16*bf16","accumulate f32","scale f32"],"group":178,"kernel":"w8a16_gemm_pipelined_m32::w8a16_gemm_pipelined_m32","family":"w8a16_gemm","point":null,"compute":"tensor_core:mma.sync.m16n8k16.bf16","numerics":"reference","where":"device","state":null},{"id":"l12.gdn.add","order":293,"layer":12,"block":"l12.gdn","block_type":"gdn","kind":"linear_attention","op":"residual_add","roles":[],"inserted":false,"precision":{"weight":null,"activation":"bf16","output":"bf16"},"steps":["compute f32"],"group":179,"kernel":"norm::residual_add_rms_norm","family":"rms_norm","point":null,"compute":"memory","numerics":"reference","where":"device","state":null},{"id":"l12.moe_ffn.post_norm","order":294,"layer":12,"block":"l12.moe_ffn","block_type":"moe_ffn","kind":"linear_attention","op":"rms_norm","roles":[],"inserted":false,"precision":{"weight":null,"activation":"f32","output":"bf16"},"steps":["compute f32"],"group":179,"kernel":"norm::residual_add_rms_norm","family":"rms_norm","point":null,"compute":"memory","numerics":"reference","where":"device","state":null},{"id":"l12.moe_ffn.router","order":295,"layer":12,"block":"l12.moe_ffn","block_type":"moe_ffn","kind":"linear_attention","op":"router","roles":[],"inserted":false,"precision":{"weight":"bf16","activation":"bf16","output":"bf16"},"steps":["act bf16","weight bf16->bf16","mma bf16*bf16","accumulate f32","scale none"],"group":180,"kernel":"gemv::dense_gemv_bf16","family":"dense_bf16","point":null,"compute":"cuda_core","numerics":"reference","where":"device","state":null},{"id":"l12.moe_ffn.top_k","order":296,"layer":12,"block":"l12.moe_ffn","block_type":"moe_ffn","kind":"linear_attention","op":"top_k","roles":[],"inserted":false,"precision":{"weight":null,"activation":"bf16","output":"f32,i32"},"steps":["score f32"],"group":181,"kernel":"moe_topk::moe_topk_softmax","family":"moe_topk","point":null,"compute":"memory","numerics":"reference","where":"device","state":null},{"id":"l12.moe_ffn.experts_gate_up","order":297,"layer":12,"block":"l12.moe_ffn","block_type":"moe_ffn","kind":"linear_attention","op":"expert_gate_up","roles":[],"inserted":false,"precision":{"weight":"fp8/block128x128","activation":"bf16","output":"bf16"},"steps":["gather bf16","act bf16","weight fp8/block128x128->bf16","mma bf16*bf16","accumulate f32","scale f32"],"group":182,"kernel":"moe_shared_expert_fused_fp8::moe_expert_gate_up_shared_fp8","family":"moe_fp8_1row","point":null,"compute":"cuda_core","numerics":"reference","where":"device","state":null},{"id":"l12.moe_ffn.shared_gate_up","order":298,"layer":12,"block":"l12.moe_ffn","block_type":"moe_ffn","kind":"linear_attention","op":"linear:shared_gate_up","roles":["shared_gate_up"],"inserted":false,"precision":{"weight":"fp8/block128x128","activation":"bf16","output":"bf16"},"steps":["act bf16","weight fp8/block128x128->bf16","mma bf16*bf16","accumulate f32","scale f32"],"group":182,"kernel":"moe_shared_expert_fused_fp8::moe_expert_gate_up_shared_fp8","family":"moe_fp8_1row","point":null,"compute":"cuda_core","numerics":"reference","where":"device","state":null},{"id":"l12.moe_ffn.experts_act","order":299,"layer":12,"block":"l12.moe_ffn","block_type":"moe_ffn","kind":"linear_attention","op":"silu_mul","roles":[],"inserted":false,"precision":{"weight":null,"activation":"bf16","output":"f32"},"steps":["compute f32"],"group":183,"kernel":"moe_shared_expert_fused_fp8::moe_expert_silu_down_shared_fp8","family":"moe_fp8_1row","point":null,"compute":"cuda_core","numerics":"reference","where":"device","state":null},{"id":"l12.moe_ffn.experts_down","order":300,"layer":12,"block":"l12.moe_ffn","block_type":"moe_ffn","kind":"linear_attention","op":"expert_down","roles":[],"inserted":false,"precision":{"weight":"fp8/block128x128","activation":"f32","output":"bf16"},"steps":["act f32","weight fp8/block128x128->f32","mma f32*f32","accumulate f32","scale f32"],"group":183,"kernel":"moe_shared_expert_fused_fp8::moe_expert_silu_down_shared_fp8","family":"moe_fp8_1row","point":null,"compute":"cuda_core","numerics":"reference","where":"device","state":null},{"id":"l12.moe_ffn.shared_act","order":301,"layer":12,"block":"l12.moe_ffn","block_type":"moe_ffn","kind":"linear_attention","op":"silu_mul","roles":[],"inserted":false,"precision":{"weight":null,"activation":"bf16","output":"f32"},"steps":["compute f32"],"group":183,"kernel":"moe_shared_expert_fused_fp8::moe_expert_silu_down_shared_fp8","family":"moe_fp8_1row","point":null,"compute":"cuda_core","numerics":"reference","where":"device","state":null},{"id":"l12.moe_ffn.shared_down","order":302,"layer":12,"block":"l12.moe_ffn","block_type":"moe_ffn","kind":"linear_attention","op":"linear:shared_down","roles":["shared_down"],"inserted":false,"precision":{"weight":"fp8/block128x128","activation":"f32","output":"bf16"},"steps":["act f32","weight fp8/block128x128->f32","mma f32*f32","accumulate f32","scale f32"],"group":183,"kernel":"moe_shared_expert_fused_fp8::moe_expert_silu_down_shared_fp8","family":"moe_fp8_1row","point":null,"compute":"cuda_core","numerics":"reference","where":"device","state":null},{"id":"l12.moe_ffn.shared_gate","order":303,"layer":12,"block":"l12.moe_ffn","block_type":"moe_ffn","kind":"linear_attention","op":"linear:shared_gate","roles":["shared_gate"],"inserted":false,"precision":{"weight":"bf16","activation":"bf16","output":"f32"},"steps":["act bf16","weight bf16->bf16","mma bf16*bf16","accumulate f32","scale none"],"group":184,"kernel":"moe_expert_gemv::moe_weighted_sum_blend","family":"moe_blend","point":{"shared_gate":"sigmoid"},"compute":"cuda_core","numerics":"reference","where":"device","state":null},{"id":"l12.moe_ffn.blend","order":304,"layer":12,"block":"l12.moe_ffn","block_type":"moe_ffn","kind":"linear_attention","op":"blend","roles":[],"inserted":false,"precision":{"weight":null,"activation":"bf16","output":"bf16"},"steps":["scatter f32","combine f32"],"group":184,"kernel":"moe_expert_gemv::moe_weighted_sum_blend","family":"moe_blend","point":{"shared_gate":"sigmoid"},"compute":"cuda_core","numerics":"reference","where":"device","state":null},{"id":"l12.moe_ffn.add","order":305,"layer":12,"block":"l12.moe_ffn","block_type":"moe_ffn","kind":"linear_attention","op":"residual_add","roles":[],"inserted":false,"precision":{"weight":null,"activation":"bf16","output":"bf16"},"steps":["compute f32"],"group":185,"kernel":"residual_add_rms_norm_exact::residual_add_rms_norm_exact","family":"rms_norm","point":null,"compute":"memory","numerics":"bit_identical(residual_add_rms_norm_exact_microtest)","where":"device","state":null},{"id":"l13.gdn.input_norm","order":306,"layer":13,"block":"l13.gdn","block_type":"gdn","kind":"linear_attention","op":"rms_norm","roles":[],"inserted":false,"precision":{"weight":null,"activation":"bf16","output":"bf16"},"steps":["compute f32"],"group":185,"kernel":"residual_add_rms_norm_exact::residual_add_rms_norm_exact","family":"rms_norm","point":null,"compute":"memory","numerics":"bit_identical(residual_add_rms_norm_exact_microtest)","where":"device","state":null},{"id":"l13.gdn.qkvz","order":307,"layer":13,"block":"l13.gdn","block_type":"gdn","kind":"linear_attention","op":"linear:qkvz","roles":["qkvz"],"inserted":false,"precision":{"weight":"fp8/block128x128","activation":"bf16","output":"bf16"},"steps":["act bf16","weight fp8/block128x128->bf16","mma bf16*bf16","accumulate f32","scale f32"],"group":186,"kernel":"w8a16_gemm_pipelined_m32::w8a16_gemm_pipelined_m32","family":"w8a16_gemm","point":null,"compute":"tensor_core:mma.sync.m16n8k16.bf16","numerics":"reference","where":"device","state":null},{"id":"l13.gdn.ba","order":308,"layer":13,"block":"l13.gdn","block_type":"gdn","kind":"linear_attention","op":"linear:ba","roles":["ba"],"inserted":false,"precision":{"weight":"bf16","activation":"bf16","output":"f32"},"steps":["act bf16","weight bf16->bf16","mma bf16*bf16","accumulate f32","scale none"],"group":187,"kernel":"ssm_preprocess::dense_gemv_ba_gates","family":"gdn_ba_gates_gemv","point":{"gate":"softplus","decay":"head"},"compute":"cuda_core","numerics":"reference","where":"device","state":null},{"id":"l13.gdn.gates","order":309,"layer":13,"block":"l13.gdn","block_type":"gdn","kind":"linear_attention","op":"gdn_gates","roles":[],"inserted":false,"precision":{"weight":null,"activation":"f32","output":"f32,f32"},"steps":["compute f32"],"group":187,"kernel":"ssm_preprocess::dense_gemv_ba_gates","family":"gdn_ba_gates_gemv","point":{"gate":"softplus","decay":"head"},"compute":"cuda_core","numerics":"reference","where":"device","state":null},{"id":"l13.gdn.conv","order":310,"layer":13,"block":"l13.gdn","block_type":"gdn","kind":"linear_attention","op":"conv1d_update","roles":[],"inserted":false,"precision":{"weight":null,"activation":"bf16","output":"f32"},"steps":["state f32","compute f32"],"group":188,"kernel":"causal_conv1d::causal_conv1d_update_l2norm_f32","family":"causal_conv1d_l2norm","point":null,"compute":"cuda_core","numerics":"reference","where":"device","state":"conv:update"},{"id":"l13.gdn.conv_ckpt","order":311,"layer":13,"block":"l13.gdn","block_type":"gdn","kind":"linear_attention","op":"state_snapshot","roles":[],"inserted":false,"precision":{"weight":null,"activation":"f32","output":null},"steps":["state f32"],"group":188,"kernel":"causal_conv1d::causal_conv1d_update_l2norm_f32","family":"causal_conv1d_l2norm","point":null,"compute":"cuda_core","numerics":"reference","where":"device","state":"conv:snapshot"},{"id":"l13.gdn.l2","order":312,"layer":13,"block":"l13.gdn","block_type":"gdn","kind":"linear_attention","op":"l2_norm","roles":[],"inserted":false,"precision":{"weight":null,"activation":"f32","output":"f32"},"steps":["compute f32"],"group":188,"kernel":"causal_conv1d::causal_conv1d_update_l2norm_f32","family":"causal_conv1d_l2norm","point":null,"compute":"cuda_core","numerics":"reference","where":"device","state":null},{"id":"l13.gdn.recur","order":313,"layer":13,"block":"l13.gdn","block_type":"gdn","kind":"linear_attention","op":"gdn_recurrence","roles":[],"inserted":false,"precision":{"weight":null,"activation":"f32","output":"f32"},"steps":["state f32","compute f32"],"group":189,"kernel":"gated_delta_rule::gated_delta_rule_decode_f32","family":"gdn_recurrence","point":{"decay":"head","gate_act":"silu"},"compute":"cuda_core","numerics":"reference","where":"device","state":"h:update"},{"id":"l13.gdn.out_norm","order":314,"layer":13,"block":"l13.gdn","block_type":"gdn","kind":"linear_attention","op":"gated_rms_norm","roles":[],"inserted":false,"precision":{"weight":null,"activation":"f32","output":"bf16"},"steps":["compute f32"],"group":190,"kernel":"norm::gated_rms_norm_f32_input","family":"gated_rms_norm","point":null,"compute":"memory","numerics":"reference","where":"device","state":null},{"id":"l13.gdn.out","order":315,"layer":13,"block":"l13.gdn","block_type":"gdn","kind":"linear_attention","op":"linear:gdn_out","roles":["gdn_out"],"inserted":false,"precision":{"weight":"fp8/block128x128","activation":"bf16","output":"bf16"},"steps":["act bf16","weight fp8/block128x128->bf16","mma bf16*bf16","accumulate f32","scale f32"],"group":191,"kernel":"w8a16_gemm_pipelined_m32::w8a16_gemm_pipelined_m32","family":"w8a16_gemm","point":null,"compute":"tensor_core:mma.sync.m16n8k16.bf16","numerics":"reference","where":"device","state":null},{"id":"l13.gdn.add","order":316,"layer":13,"block":"l13.gdn","block_type":"gdn","kind":"linear_attention","op":"residual_add","roles":[],"inserted":false,"precision":{"weight":null,"activation":"bf16","output":"bf16"},"steps":["compute f32"],"group":192,"kernel":"norm::residual_add_rms_norm","family":"rms_norm","point":null,"compute":"memory","numerics":"reference","where":"device","state":null},{"id":"l13.moe_ffn.post_norm","order":317,"layer":13,"block":"l13.moe_ffn","block_type":"moe_ffn","kind":"linear_attention","op":"rms_norm","roles":[],"inserted":false,"precision":{"weight":null,"activation":"f32","output":"bf16"},"steps":["compute f32"],"group":192,"kernel":"norm::residual_add_rms_norm","family":"rms_norm","point":null,"compute":"memory","numerics":"reference","where":"device","state":null},{"id":"l13.moe_ffn.router","order":318,"layer":13,"block":"l13.moe_ffn","block_type":"moe_ffn","kind":"linear_attention","op":"router","roles":[],"inserted":false,"precision":{"weight":"bf16","activation":"bf16","output":"bf16"},"steps":["act bf16","weight bf16->bf16","mma bf16*bf16","accumulate f32","scale none"],"group":193,"kernel":"gemv::dense_gemv_bf16","family":"dense_bf16","point":null,"compute":"cuda_core","numerics":"reference","where":"device","state":null},{"id":"l13.moe_ffn.top_k","order":319,"layer":13,"block":"l13.moe_ffn","block_type":"moe_ffn","kind":"linear_attention","op":"top_k","roles":[],"inserted":false,"precision":{"weight":null,"activation":"bf16","output":"f32,i32"},"steps":["score f32"],"group":194,"kernel":"moe_topk::moe_topk_softmax","family":"moe_topk","point":null,"compute":"memory","numerics":"reference","where":"device","state":null},{"id":"l13.moe_ffn.experts_gate_up","order":320,"layer":13,"block":"l13.moe_ffn","block_type":"moe_ffn","kind":"linear_attention","op":"expert_gate_up","roles":[],"inserted":false,"precision":{"weight":"fp8/block128x128","activation":"bf16","output":"bf16"},"steps":["gather bf16","act bf16","weight fp8/block128x128->bf16","mma bf16*bf16","accumulate f32","scale f32"],"group":195,"kernel":"moe_shared_expert_fused_fp8::moe_expert_gate_up_shared_fp8","family":"moe_fp8_1row","point":null,"compute":"cuda_core","numerics":"reference","where":"device","state":null},{"id":"l13.moe_ffn.shared_gate_up","order":321,"layer":13,"block":"l13.moe_ffn","block_type":"moe_ffn","kind":"linear_attention","op":"linear:shared_gate_up","roles":["shared_gate_up"],"inserted":false,"precision":{"weight":"fp8/block128x128","activation":"bf16","output":"bf16"},"steps":["act bf16","weight fp8/block128x128->bf16","mma bf16*bf16","accumulate f32","scale f32"],"group":195,"kernel":"moe_shared_expert_fused_fp8::moe_expert_gate_up_shared_fp8","family":"moe_fp8_1row","point":null,"compute":"cuda_core","numerics":"reference","where":"device","state":null},{"id":"l13.moe_ffn.experts_act","order":322,"layer":13,"block":"l13.moe_ffn","block_type":"moe_ffn","kind":"linear_attention","op":"silu_mul","roles":[],"inserted":false,"precision":{"weight":null,"activation":"bf16","output":"f32"},"steps":["compute f32"],"group":196,"kernel":"moe_shared_expert_fused_fp8::moe_expert_silu_down_shared_fp8","family":"moe_fp8_1row","point":null,"compute":"cuda_core","numerics":"reference","where":"device","state":null},{"id":"l13.moe_ffn.experts_down","order":323,"layer":13,"block":"l13.moe_ffn","block_type":"moe_ffn","kind":"linear_attention","op":"expert_down","roles":[],"inserted":false,"precision":{"weight":"fp8/block128x128","activation":"f32","output":"bf16"},"steps":["act f32","weight fp8/block128x128->f32","mma f32*f32","accumulate f32","scale f32"],"group":196,"kernel":"moe_shared_expert_fused_fp8::moe_expert_silu_down_shared_fp8","family":"moe_fp8_1row","point":null,"compute":"cuda_core","numerics":"reference","where":"device","state":null},{"id":"l13.moe_ffn.shared_act","order":324,"layer":13,"block":"l13.moe_ffn","block_type":"moe_ffn","kind":"linear_attention","op":"silu_mul","roles":[],"inserted":false,"precision":{"weight":null,"activation":"bf16","output":"f32"},"steps":["compute f32"],"group":196,"kernel":"moe_shared_expert_fused_fp8::moe_expert_silu_down_shared_fp8","family":"moe_fp8_1row","point":null,"compute":"cuda_core","numerics":"reference","where":"device","state":null},{"id":"l13.moe_ffn.shared_down","order":325,"layer":13,"block":"l13.moe_ffn","block_type":"moe_ffn","kind":"linear_attention","op":"linear:shared_down","roles":["shared_down"],"inserted":false,"precision":{"weight":"fp8/block128x128","activation":"f32","output":"bf16"},"steps":["act f32","weight fp8/block128x128->f32","mma f32*f32","accumulate f32","scale f32"],"group":196,"kernel":"moe_shared_expert_fused_fp8::moe_expert_silu_down_shared_fp8","family":"moe_fp8_1row","point":null,"compute":"cuda_core","numerics":"reference","where":"device","state":null},{"id":"l13.moe_ffn.shared_gate","order":326,"layer":13,"block":"l13.moe_ffn","block_type":"moe_ffn","kind":"linear_attention","op":"linear:shared_gate","roles":["shared_gate"],"inserted":false,"precision":{"weight":"bf16","activation":"bf16","output":"f32"},"steps":["act bf16","weight bf16->bf16","mma bf16*bf16","accumulate f32","scale none"],"group":197,"kernel":"moe_expert_gemv::moe_weighted_sum_blend","family":"moe_blend","point":{"shared_gate":"sigmoid"},"compute":"cuda_core","numerics":"reference","where":"device","state":null},{"id":"l13.moe_ffn.blend","order":327,"layer":13,"block":"l13.moe_ffn","block_type":"moe_ffn","kind":"linear_attention","op":"blend","roles":[],"inserted":false,"precision":{"weight":null,"activation":"bf16","output":"bf16"},"steps":["scatter f32","combine f32"],"group":197,"kernel":"moe_expert_gemv::moe_weighted_sum_blend","family":"moe_blend","point":{"shared_gate":"sigmoid"},"compute":"cuda_core","numerics":"reference","where":"device","state":null},{"id":"l13.moe_ffn.add","order":328,"layer":13,"block":"l13.moe_ffn","block_type":"moe_ffn","kind":"linear_attention","op":"residual_add","roles":[],"inserted":false,"precision":{"weight":null,"activation":"bf16","output":"bf16"},"steps":["compute f32"],"group":198,"kernel":"residual_add_rms_norm_exact::residual_add_rms_norm_exact","family":"rms_norm","point":null,"compute":"memory","numerics":"bit_identical(residual_add_rms_norm_exact_microtest)","where":"device","state":null},{"id":"l14.gdn.input_norm","order":329,"layer":14,"block":"l14.gdn","block_type":"gdn","kind":"linear_attention","op":"rms_norm","roles":[],"inserted":false,"precision":{"weight":null,"activation":"bf16","output":"bf16"},"steps":["compute f32"],"group":198,"kernel":"residual_add_rms_norm_exact::residual_add_rms_norm_exact","family":"rms_norm","point":null,"compute":"memory","numerics":"bit_identical(residual_add_rms_norm_exact_microtest)","where":"device","state":null},{"id":"l14.gdn.qkvz","order":330,"layer":14,"block":"l14.gdn","block_type":"gdn","kind":"linear_attention","op":"linear:qkvz","roles":["qkvz"],"inserted":false,"precision":{"weight":"fp8/block128x128","activation":"bf16","output":"bf16"},"steps":["act bf16","weight fp8/block128x128->bf16","mma bf16*bf16","accumulate f32","scale f32"],"group":199,"kernel":"w8a16_gemm_pipelined_m32::w8a16_gemm_pipelined_m32","family":"w8a16_gemm","point":null,"compute":"tensor_core:mma.sync.m16n8k16.bf16","numerics":"reference","where":"device","state":null},{"id":"l14.gdn.ba","order":331,"layer":14,"block":"l14.gdn","block_type":"gdn","kind":"linear_attention","op":"linear:ba","roles":["ba"],"inserted":false,"precision":{"weight":"bf16","activation":"bf16","output":"f32"},"steps":["act bf16","weight bf16->bf16","mma bf16*bf16","accumulate f32","scale none"],"group":200,"kernel":"ssm_preprocess::dense_gemv_ba_gates","family":"gdn_ba_gates_gemv","point":{"gate":"softplus","decay":"head"},"compute":"cuda_core","numerics":"reference","where":"device","state":null},{"id":"l14.gdn.gates","order":332,"layer":14,"block":"l14.gdn","block_type":"gdn","kind":"linear_attention","op":"gdn_gates","roles":[],"inserted":false,"precision":{"weight":null,"activation":"f32","output":"f32,f32"},"steps":["compute f32"],"group":200,"kernel":"ssm_preprocess::dense_gemv_ba_gates","family":"gdn_ba_gates_gemv","point":{"gate":"softplus","decay":"head"},"compute":"cuda_core","numerics":"reference","where":"device","state":null},{"id":"l14.gdn.conv","order":333,"layer":14,"block":"l14.gdn","block_type":"gdn","kind":"linear_attention","op":"conv1d_update","roles":[],"inserted":false,"precision":{"weight":null,"activation":"bf16","output":"f32"},"steps":["state f32","compute f32"],"group":201,"kernel":"causal_conv1d::causal_conv1d_update_l2norm_f32","family":"causal_conv1d_l2norm","point":null,"compute":"cuda_core","numerics":"reference","where":"device","state":"conv:update"},{"id":"l14.gdn.conv_ckpt","order":334,"layer":14,"block":"l14.gdn","block_type":"gdn","kind":"linear_attention","op":"state_snapshot","roles":[],"inserted":false,"precision":{"weight":null,"activation":"f32","output":null},"steps":["state f32"],"group":201,"kernel":"causal_conv1d::causal_conv1d_update_l2norm_f32","family":"causal_conv1d_l2norm","point":null,"compute":"cuda_core","numerics":"reference","where":"device","state":"conv:snapshot"},{"id":"l14.gdn.l2","order":335,"layer":14,"block":"l14.gdn","block_type":"gdn","kind":"linear_attention","op":"l2_norm","roles":[],"inserted":false,"precision":{"weight":null,"activation":"f32","output":"f32"},"steps":["compute f32"],"group":201,"kernel":"causal_conv1d::causal_conv1d_update_l2norm_f32","family":"causal_conv1d_l2norm","point":null,"compute":"cuda_core","numerics":"reference","where":"device","state":null},{"id":"l14.gdn.recur","order":336,"layer":14,"block":"l14.gdn","block_type":"gdn","kind":"linear_attention","op":"gdn_recurrence","roles":[],"inserted":false,"precision":{"weight":null,"activation":"f32","output":"f32"},"steps":["state f32","compute f32"],"group":202,"kernel":"gated_delta_rule::gated_delta_rule_decode_f32","family":"gdn_recurrence","point":{"decay":"head","gate_act":"silu"},"compute":"cuda_core","numerics":"reference","where":"device","state":"h:update"},{"id":"l14.gdn.out_norm","order":337,"layer":14,"block":"l14.gdn","block_type":"gdn","kind":"linear_attention","op":"gated_rms_norm","roles":[],"inserted":false,"precision":{"weight":null,"activation":"f32","output":"bf16"},"steps":["compute f32"],"group":203,"kernel":"norm::gated_rms_norm_f32_input","family":"gated_rms_norm","point":null,"compute":"memory","numerics":"reference","where":"device","state":null},{"id":"l14.gdn.out","order":338,"layer":14,"block":"l14.gdn","block_type":"gdn","kind":"linear_attention","op":"linear:gdn_out","roles":["gdn_out"],"inserted":false,"precision":{"weight":"fp8/block128x128","activation":"bf16","output":"bf16"},"steps":["act bf16","weight fp8/block128x128->bf16","mma bf16*bf16","accumulate f32","scale f32"],"group":204,"kernel":"w8a16_gemm_pipelined_m32::w8a16_gemm_pipelined_m32","family":"w8a16_gemm","point":null,"compute":"tensor_core:mma.sync.m16n8k16.bf16","numerics":"reference","where":"device","state":null},{"id":"l14.gdn.add","order":339,"layer":14,"block":"l14.gdn","block_type":"gdn","kind":"linear_attention","op":"residual_add","roles":[],"inserted":false,"precision":{"weight":null,"activation":"bf16","output":"bf16"},"steps":["compute f32"],"group":205,"kernel":"norm::residual_add_rms_norm","family":"rms_norm","point":null,"compute":"memory","numerics":"reference","where":"device","state":null},{"id":"l14.moe_ffn.post_norm","order":340,"layer":14,"block":"l14.moe_ffn","block_type":"moe_ffn","kind":"linear_attention","op":"rms_norm","roles":[],"inserted":false,"precision":{"weight":null,"activation":"f32","output":"bf16"},"steps":["compute f32"],"group":205,"kernel":"norm::residual_add_rms_norm","family":"rms_norm","point":null,"compute":"memory","numerics":"reference","where":"device","state":null},{"id":"l14.moe_ffn.router","order":341,"layer":14,"block":"l14.moe_ffn","block_type":"moe_ffn","kind":"linear_attention","op":"router","roles":[],"inserted":false,"precision":{"weight":"bf16","activation":"bf16","output":"bf16"},"steps":["act bf16","weight bf16->bf16","mma bf16*bf16","accumulate f32","scale none"],"group":206,"kernel":"gemv::dense_gemv_bf16","family":"dense_bf16","point":null,"compute":"cuda_core","numerics":"reference","where":"device","state":null},{"id":"l14.moe_ffn.top_k","order":342,"layer":14,"block":"l14.moe_ffn","block_type":"moe_ffn","kind":"linear_attention","op":"top_k","roles":[],"inserted":false,"precision":{"weight":null,"activation":"bf16","output":"f32,i32"},"steps":["score f32"],"group":207,"kernel":"moe_topk::moe_topk_softmax","family":"moe_topk","point":null,"compute":"memory","numerics":"reference","where":"device","state":null},{"id":"l14.moe_ffn.experts_gate_up","order":343,"layer":14,"block":"l14.moe_ffn","block_type":"moe_ffn","kind":"linear_attention","op":"expert_gate_up","roles":[],"inserted":false,"precision":{"weight":"fp8/block128x128","activation":"bf16","output":"bf16"},"steps":["gather bf16","act bf16","weight fp8/block128x128->bf16","mma bf16*bf16","accumulate f32","scale f32"],"group":208,"kernel":"moe_shared_expert_fused_fp8::moe_expert_gate_up_shared_fp8","family":"moe_fp8_1row","point":null,"compute":"cuda_core","numerics":"reference","where":"device","state":null},{"id":"l14.moe_ffn.shared_gate_up","order":344,"layer":14,"block":"l14.moe_ffn","block_type":"moe_ffn","kind":"linear_attention","op":"linear:shared_gate_up","roles":["shared_gate_up"],"inserted":false,"precision":{"weight":"fp8/block128x128","activation":"bf16","output":"bf16"},"steps":["act bf16","weight fp8/block128x128->bf16","mma bf16*bf16","accumulate f32","scale f32"],"group":208,"kernel":"moe_shared_expert_fused_fp8::moe_expert_gate_up_shared_fp8","family":"moe_fp8_1row","point":null,"compute":"cuda_core","numerics":"reference","where":"device","state":null},{"id":"l14.moe_ffn.experts_act","order":345,"layer":14,"block":"l14.moe_ffn","block_type":"moe_ffn","kind":"linear_attention","op":"silu_mul","roles":[],"inserted":false,"precision":{"weight":null,"activation":"bf16","output":"f32"},"steps":["compute f32"],"group":209,"kernel":"moe_shared_expert_fused_fp8::moe_expert_silu_down_shared_fp8","family":"moe_fp8_1row","point":null,"compute":"cuda_core","numerics":"reference","where":"device","state":null},{"id":"l14.moe_ffn.experts_down","order":346,"layer":14,"block":"l14.moe_ffn","block_type":"moe_ffn","kind":"linear_attention","op":"expert_down","roles":[],"inserted":false,"precision":{"weight":"fp8/block128x128","activation":"f32","output":"bf16"},"steps":["act f32","weight fp8/block128x128->f32","mma f32*f32","accumulate f32","scale f32"],"group":209,"kernel":"moe_shared_expert_fused_fp8::moe_expert_silu_down_shared_fp8","family":"moe_fp8_1row","point":null,"compute":"cuda_core","numerics":"reference","where":"device","state":null},{"id":"l14.moe_ffn.shared_act","order":347,"layer":14,"block":"l14.moe_ffn","block_type":"moe_ffn","kind":"linear_attention","op":"silu_mul","roles":[],"inserted":false,"precision":{"weight":null,"activation":"bf16","output":"f32"},"steps":["compute f32"],"group":209,"kernel":"moe_shared_expert_fused_fp8::moe_expert_silu_down_shared_fp8","family":"moe_fp8_1row","point":null,"compute":"cuda_core","numerics":"reference","where":"device","state":null},{"id":"l14.moe_ffn.shared_down","order":348,"layer":14,"block":"l14.moe_ffn","block_type":"moe_ffn","kind":"linear_attention","op":"linear:shared_down","roles":["shared_down"],"inserted":false,"precision":{"weight":"fp8/block128x128","activation":"f32","output":"bf16"},"steps":["act f32","weight fp8/block128x128->f32","mma f32*f32","accumulate f32","scale f32"],"group":209,"kernel":"moe_shared_expert_fused_fp8::moe_expert_silu_down_shared_fp8","family":"moe_fp8_1row","point":null,"compute":"cuda_core","numerics":"reference","where":"device","state":null},{"id":"l14.moe_ffn.shared_gate","order":349,"layer":14,"block":"l14.moe_ffn","block_type":"moe_ffn","kind":"linear_attention","op":"linear:shared_gate","roles":["shared_gate"],"inserted":false,"precision":{"weight":"bf16","activation":"bf16","output":"f32"},"steps":["act bf16","weight bf16->bf16","mma bf16*bf16","accumulate f32","scale none"],"group":210,"kernel":"moe_expert_gemv::moe_weighted_sum_blend","family":"moe_blend","point":{"shared_gate":"sigmoid"},"compute":"cuda_core","numerics":"reference","where":"device","state":null},{"id":"l14.moe_ffn.blend","order":350,"layer":14,"block":"l14.moe_ffn","block_type":"moe_ffn","kind":"linear_attention","op":"blend","roles":[],"inserted":false,"precision":{"weight":null,"activation":"bf16","output":"bf16"},"steps":["scatter f32","combine f32"],"group":210,"kernel":"moe_expert_gemv::moe_weighted_sum_blend","family":"moe_blend","point":{"shared_gate":"sigmoid"},"compute":"cuda_core","numerics":"reference","where":"device","state":null},{"id":"l14.moe_ffn.add","order":351,"layer":14,"block":"l14.moe_ffn","block_type":"moe_ffn","kind":"linear_attention","op":"residual_add","roles":[],"inserted":false,"precision":{"weight":null,"activation":"bf16","output":"bf16"},"steps":["compute f32"],"group":211,"kernel":"residual_add_rms_norm_exact::residual_add_rms_norm_exact","family":"rms_norm","point":null,"compute":"memory","numerics":"bit_identical(residual_add_rms_norm_exact_microtest)","where":"device","state":null},{"id":"l15.attn.input_norm","order":352,"layer":15,"block":"l15.attn","block_type":"attn","kind":"full_attention","op":"rms_norm","roles":[],"inserted":false,"precision":{"weight":null,"activation":"bf16","output":"bf16"},"steps":["compute f32"],"group":211,"kernel":"residual_add_rms_norm_exact::residual_add_rms_norm_exact","family":"rms_norm","point":null,"compute":"memory","numerics":"bit_identical(residual_add_rms_norm_exact_microtest)","where":"device","state":null},{"id":"l15.attn.q","order":353,"layer":15,"block":"l15.attn","block_type":"attn","kind":"full_attention","op":"linear:q","roles":["q"],"inserted":false,"precision":{"weight":"fp8/block128x128","activation":"bf16","output":"bf16"},"steps":["act bf16","weight fp8/block128x128->bf16","mma bf16*bf16","accumulate f32","scale f32"],"group":212,"kernel":"w8a16_gemm_pipelined_m32::w8a16_gemm_pipelined_m32","family":"w8a16_gemm","point":null,"compute":"tensor_core:mma.sync.m16n8k16.bf16","numerics":"reference","where":"device","state":null},{"id":"l15.attn.q_split","order":354,"layer":15,"block":"l15.attn","block_type":"attn","kind":"full_attention","op":"split","roles":[],"inserted":false,"precision":{"weight":null,"activation":"bf16","output":"bf16,bf16"},"steps":["move bf16"],"group":213,"kernel":"ssm_preprocess::deinterleave_qg","family":"deinterleave_qg","point":{"layout":"interleaved"},"compute":"memory","numerics":"reference","where":"device","state":null},{"id":"l15.attn.k","order":355,"layer":15,"block":"l15.attn","block_type":"attn","kind":"full_attention","op":"linear:k","roles":["k"],"inserted":false,"precision":{"weight":"fp8/block128x128","activation":"bf16","output":"bf16"},"steps":["act bf16","weight fp8/block128x128->bf16","mma bf16*bf16","accumulate f32","scale f32"],"group":214,"kernel":"w8a16_gemm_pipelined_m32::w8a16_gemm_pipelined_m32","family":"w8a16_gemm","point":null,"compute":"tensor_core:mma.sync.m16n8k16.bf16","numerics":"reference","where":"device","state":null},{"id":"l15.attn.v","order":356,"layer":15,"block":"l15.attn","block_type":"attn","kind":"full_attention","op":"linear:v","roles":["v"],"inserted":false,"precision":{"weight":"fp8/block128x128","activation":"bf16","output":"bf16"},"steps":["act bf16","weight fp8/block128x128->bf16","mma bf16*bf16","accumulate f32","scale f32"],"group":215,"kernel":"w8a16_gemm_pipelined_m32::w8a16_gemm_pipelined_m32","family":"w8a16_gemm","point":null,"compute":"tensor_core:mma.sync.m16n8k16.bf16","numerics":"reference","where":"device","state":null},{"id":"l15.attn.q_norm","order":357,"layer":15,"block":"l15.attn","block_type":"attn","kind":"full_attention","op":"qk_norm","roles":[],"inserted":false,"precision":{"weight":null,"activation":"bf16","output":"bf16"},"steps":["compute f32"],"group":216,"kernel":"norm::rms_norm","family":"rms_norm","point":null,"compute":"memory","numerics":"reference","where":"device","state":null},{"id":"l15.attn.k_norm","order":358,"layer":15,"block":"l15.attn","block_type":"attn","kind":"full_attention","op":"qk_norm","roles":[],"inserted":false,"precision":{"weight":null,"activation":"bf16","output":"bf16"},"steps":["compute f32"],"group":217,"kernel":"norm::rms_norm","family":"rms_norm","point":null,"compute":"memory","numerics":"reference","where":"device","state":null},{"id":"l15.attn.rope","order":359,"layer":15,"block":"l15.attn","block_type":"attn","kind":"full_attention","op":"rope","roles":[],"inserted":false,"precision":{"weight":null,"activation":"bf16","output":"bf16,bf16"},"steps":["compute f32"],"group":218,"kernel":"rope_mrope_interleaved::rope_forward_mrope_interleaved","family":"rope","point":null,"compute":"memory","numerics":"reference","where":"device","state":null},{"id":"l15.attn.kv_write","order":360,"layer":15,"block":"l15.attn","block_type":"attn","kind":"full_attention","op":"kv_write","roles":[],"inserted":false,"precision":{"weight":null,"activation":"bf16","output":null},"steps":["cache bf16"],"group":219,"kernel":"reshape_and_cache::reshape_and_cache_flash","family":"kv_write","point":null,"compute":"memory","numerics":"reference","where":"device","state":"k:write,v:write"},{"id":"l15.attn.attend","order":361,"layer":15,"block":"l15.attn","block_type":"attn","kind":"full_attention","op":"paged_attention","roles":[],"inserted":false,"precision":{"weight":null,"activation":"bf16","output":"bf16"},"steps":["cache bf16","scores f32","softmax f32","accumulate f32"],"group":220,"kernel":"paged_decode::paged_decode_attn","family":"paged_decode_attn","point":null,"compute":"cuda_core","numerics":"reference","where":"device","state":"k:read,v:read"},{"id":"l15.attn.gate_mul","order":362,"layer":15,"block":"l15.attn","block_type":"attn","kind":"full_attention","op":"sigmoid_gate_mul","roles":[],"inserted":false,"precision":{"weight":null,"activation":"bf16","output":"bf16"},"steps":["compute f32"],"group":221,"kernel":"residual_add::sigmoid_gate_mul","family":"sigmoid_gate_mul","point":null,"compute":"memory","numerics":"reference","where":"device","state":null},{"id":"l15.attn.o","order":363,"layer":15,"block":"l15.attn","block_type":"attn","kind":"full_attention","op":"linear:o","roles":["o"],"inserted":false,"precision":{"weight":"fp8/block128x128","activation":"bf16","output":"bf16"},"steps":["act bf16","weight fp8/block128x128->bf16","mma bf16*bf16","accumulate f32","scale f32"],"group":222,"kernel":"w8a16_gemm_pipelined_m32::w8a16_gemm_pipelined_m32","family":"w8a16_gemm","point":null,"compute":"tensor_core:mma.sync.m16n8k16.bf16","numerics":"reference","where":"device","state":null},{"id":"l15.attn.add","order":364,"layer":15,"block":"l15.attn","block_type":"attn","kind":"full_attention","op":"residual_add","roles":[],"inserted":false,"precision":{"weight":null,"activation":"bf16","output":"bf16"},"steps":["compute f32"],"group":223,"kernel":"norm::residual_add_rms_norm","family":"rms_norm","point":null,"compute":"memory","numerics":"reference","where":"device","state":null},{"id":"l15.moe_ffn.post_norm","order":365,"layer":15,"block":"l15.moe_ffn","block_type":"moe_ffn","kind":"full_attention","op":"rms_norm","roles":[],"inserted":false,"precision":{"weight":null,"activation":"f32","output":"bf16"},"steps":["compute f32"],"group":223,"kernel":"norm::residual_add_rms_norm","family":"rms_norm","point":null,"compute":"memory","numerics":"reference","where":"device","state":null},{"id":"l15.moe_ffn.router","order":366,"layer":15,"block":"l15.moe_ffn","block_type":"moe_ffn","kind":"full_attention","op":"router","roles":[],"inserted":false,"precision":{"weight":"bf16","activation":"bf16","output":"bf16"},"steps":["act bf16","weight bf16->bf16","mma bf16*bf16","accumulate f32","scale none"],"group":224,"kernel":"gemv::dense_gemv_bf16","family":"dense_bf16","point":null,"compute":"cuda_core","numerics":"reference","where":"device","state":null},{"id":"l15.moe_ffn.top_k","order":367,"layer":15,"block":"l15.moe_ffn","block_type":"moe_ffn","kind":"full_attention","op":"top_k","roles":[],"inserted":false,"precision":{"weight":null,"activation":"bf16","output":"f32,i32"},"steps":["score f32"],"group":225,"kernel":"moe_topk::moe_topk_softmax","family":"moe_topk","point":null,"compute":"memory","numerics":"reference","where":"device","state":null},{"id":"l15.moe_ffn.experts_gate_up","order":368,"layer":15,"block":"l15.moe_ffn","block_type":"moe_ffn","kind":"full_attention","op":"expert_gate_up","roles":[],"inserted":false,"precision":{"weight":"fp8/block128x128","activation":"bf16","output":"bf16"},"steps":["gather bf16","act bf16","weight fp8/block128x128->bf16","mma bf16*bf16","accumulate f32","scale f32"],"group":226,"kernel":"moe_shared_expert_fused_fp8::moe_expert_gate_up_shared_fp8","family":"moe_fp8_1row","point":null,"compute":"cuda_core","numerics":"reference","where":"device","state":null},{"id":"l15.moe_ffn.shared_gate_up","order":369,"layer":15,"block":"l15.moe_ffn","block_type":"moe_ffn","kind":"full_attention","op":"linear:shared_gate_up","roles":["shared_gate_up"],"inserted":false,"precision":{"weight":"fp8/block128x128","activation":"bf16","output":"bf16"},"steps":["act bf16","weight fp8/block128x128->bf16","mma bf16*bf16","accumulate f32","scale f32"],"group":226,"kernel":"moe_shared_expert_fused_fp8::moe_expert_gate_up_shared_fp8","family":"moe_fp8_1row","point":null,"compute":"cuda_core","numerics":"reference","where":"device","state":null},{"id":"l15.moe_ffn.experts_act","order":370,"layer":15,"block":"l15.moe_ffn","block_type":"moe_ffn","kind":"full_attention","op":"silu_mul","roles":[],"inserted":false,"precision":{"weight":null,"activation":"bf16","output":"f32"},"steps":["compute f32"],"group":227,"kernel":"moe_shared_expert_fused_fp8::moe_expert_silu_down_shared_fp8","family":"moe_fp8_1row","point":null,"compute":"cuda_core","numerics":"reference","where":"device","state":null},{"id":"l15.moe_ffn.experts_down","order":371,"layer":15,"block":"l15.moe_ffn","block_type":"moe_ffn","kind":"full_attention","op":"expert_down","roles":[],"inserted":false,"precision":{"weight":"fp8/block128x128","activation":"f32","output":"bf16"},"steps":["act f32","weight fp8/block128x128->f32","mma f32*f32","accumulate f32","scale f32"],"group":227,"kernel":"moe_shared_expert_fused_fp8::moe_expert_silu_down_shared_fp8","family":"moe_fp8_1row","point":null,"compute":"cuda_core","numerics":"reference","where":"device","state":null},{"id":"l15.moe_ffn.shared_act","order":372,"layer":15,"block":"l15.moe_ffn","block_type":"moe_ffn","kind":"full_attention","op":"silu_mul","roles":[],"inserted":false,"precision":{"weight":null,"activation":"bf16","output":"f32"},"steps":["compute f32"],"group":227,"kernel":"moe_shared_expert_fused_fp8::moe_expert_silu_down_shared_fp8","family":"moe_fp8_1row","point":null,"compute":"cuda_core","numerics":"reference","where":"device","state":null},{"id":"l15.moe_ffn.shared_down","order":373,"layer":15,"block":"l15.moe_ffn","block_type":"moe_ffn","kind":"full_attention","op":"linear:shared_down","roles":["shared_down"],"inserted":false,"precision":{"weight":"fp8/block128x128","activation":"f32","output":"bf16"},"steps":["act f32","weight fp8/block128x128->f32","mma f32*f32","accumulate f32","scale f32"],"group":227,"kernel":"moe_shared_expert_fused_fp8::moe_expert_silu_down_shared_fp8","family":"moe_fp8_1row","point":null,"compute":"cuda_core","numerics":"reference","where":"device","state":null},{"id":"l15.moe_ffn.shared_gate","order":374,"layer":15,"block":"l15.moe_ffn","block_type":"moe_ffn","kind":"full_attention","op":"linear:shared_gate","roles":["shared_gate"],"inserted":false,"precision":{"weight":"bf16","activation":"bf16","output":"f32"},"steps":["act bf16","weight bf16->bf16","mma bf16*bf16","accumulate f32","scale none"],"group":228,"kernel":"moe_expert_gemv::moe_weighted_sum_blend","family":"moe_blend","point":{"shared_gate":"sigmoid"},"compute":"cuda_core","numerics":"reference","where":"device","state":null},{"id":"l15.moe_ffn.blend","order":375,"layer":15,"block":"l15.moe_ffn","block_type":"moe_ffn","kind":"full_attention","op":"blend","roles":[],"inserted":false,"precision":{"weight":null,"activation":"bf16","output":"bf16"},"steps":["scatter f32","combine f32"],"group":228,"kernel":"moe_expert_gemv::moe_weighted_sum_blend","family":"moe_blend","point":{"shared_gate":"sigmoid"},"compute":"cuda_core","numerics":"reference","where":"device","state":null},{"id":"l15.moe_ffn.add","order":376,"layer":15,"block":"l15.moe_ffn","block_type":"moe_ffn","kind":"full_attention","op":"residual_add","roles":[],"inserted":false,"precision":{"weight":null,"activation":"bf16","output":"bf16"},"steps":["compute f32"],"group":229,"kernel":"residual_add_rms_norm_exact::residual_add_rms_norm_exact","family":"rms_norm","point":null,"compute":"memory","numerics":"bit_identical(residual_add_rms_norm_exact_microtest)","where":"device","state":null},{"id":"l16.gdn.input_norm","order":377,"layer":16,"block":"l16.gdn","block_type":"gdn","kind":"linear_attention","op":"rms_norm","roles":[],"inserted":false,"precision":{"weight":null,"activation":"bf16","output":"bf16"},"steps":["compute f32"],"group":229,"kernel":"residual_add_rms_norm_exact::residual_add_rms_norm_exact","family":"rms_norm","point":null,"compute":"memory","numerics":"bit_identical(residual_add_rms_norm_exact_microtest)","where":"device","state":null},{"id":"l16.gdn.qkvz","order":378,"layer":16,"block":"l16.gdn","block_type":"gdn","kind":"linear_attention","op":"linear:qkvz","roles":["qkvz"],"inserted":false,"precision":{"weight":"fp8/block128x128","activation":"bf16","output":"bf16"},"steps":["act bf16","weight fp8/block128x128->bf16","mma bf16*bf16","accumulate f32","scale f32"],"group":230,"kernel":"w8a16_gemm_pipelined_m32::w8a16_gemm_pipelined_m32","family":"w8a16_gemm","point":null,"compute":"tensor_core:mma.sync.m16n8k16.bf16","numerics":"reference","where":"device","state":null},{"id":"l16.gdn.ba","order":379,"layer":16,"block":"l16.gdn","block_type":"gdn","kind":"linear_attention","op":"linear:ba","roles":["ba"],"inserted":false,"precision":{"weight":"bf16","activation":"bf16","output":"f32"},"steps":["act bf16","weight bf16->bf16","mma bf16*bf16","accumulate f32","scale none"],"group":231,"kernel":"ssm_preprocess::dense_gemv_ba_gates","family":"gdn_ba_gates_gemv","point":{"gate":"softplus","decay":"head"},"compute":"cuda_core","numerics":"reference","where":"device","state":null},{"id":"l16.gdn.gates","order":380,"layer":16,"block":"l16.gdn","block_type":"gdn","kind":"linear_attention","op":"gdn_gates","roles":[],"inserted":false,"precision":{"weight":null,"activation":"f32","output":"f32,f32"},"steps":["compute f32"],"group":231,"kernel":"ssm_preprocess::dense_gemv_ba_gates","family":"gdn_ba_gates_gemv","point":{"gate":"softplus","decay":"head"},"compute":"cuda_core","numerics":"reference","where":"device","state":null},{"id":"l16.gdn.conv","order":381,"layer":16,"block":"l16.gdn","block_type":"gdn","kind":"linear_attention","op":"conv1d_update","roles":[],"inserted":false,"precision":{"weight":null,"activation":"bf16","output":"f32"},"steps":["state f32","compute f32"],"group":232,"kernel":"causal_conv1d::causal_conv1d_update_l2norm_f32","family":"causal_conv1d_l2norm","point":null,"compute":"cuda_core","numerics":"reference","where":"device","state":"conv:update"},{"id":"l16.gdn.conv_ckpt","order":382,"layer":16,"block":"l16.gdn","block_type":"gdn","kind":"linear_attention","op":"state_snapshot","roles":[],"inserted":false,"precision":{"weight":null,"activation":"f32","output":null},"steps":["state f32"],"group":232,"kernel":"causal_conv1d::causal_conv1d_update_l2norm_f32","family":"causal_conv1d_l2norm","point":null,"compute":"cuda_core","numerics":"reference","where":"device","state":"conv:snapshot"},{"id":"l16.gdn.l2","order":383,"layer":16,"block":"l16.gdn","block_type":"gdn","kind":"linear_attention","op":"l2_norm","roles":[],"inserted":false,"precision":{"weight":null,"activation":"f32","output":"f32"},"steps":["compute f32"],"group":232,"kernel":"causal_conv1d::causal_conv1d_update_l2norm_f32","family":"causal_conv1d_l2norm","point":null,"compute":"cuda_core","numerics":"reference","where":"device","state":null},{"id":"l16.gdn.recur","order":384,"layer":16,"block":"l16.gdn","block_type":"gdn","kind":"linear_attention","op":"gdn_recurrence","roles":[],"inserted":false,"precision":{"weight":null,"activation":"f32","output":"f32"},"steps":["state f32","compute f32"],"group":233,"kernel":"gated_delta_rule::gated_delta_rule_decode_f32","family":"gdn_recurrence","point":{"decay":"head","gate_act":"silu"},"compute":"cuda_core","numerics":"reference","where":"device","state":"h:update"},{"id":"l16.gdn.out_norm","order":385,"layer":16,"block":"l16.gdn","block_type":"gdn","kind":"linear_attention","op":"gated_rms_norm","roles":[],"inserted":false,"precision":{"weight":null,"activation":"f32","output":"bf16"},"steps":["compute f32"],"group":234,"kernel":"norm::gated_rms_norm_f32_input","family":"gated_rms_norm","point":null,"compute":"memory","numerics":"reference","where":"device","state":null},{"id":"l16.gdn.out","order":386,"layer":16,"block":"l16.gdn","block_type":"gdn","kind":"linear_attention","op":"linear:gdn_out","roles":["gdn_out"],"inserted":false,"precision":{"weight":"fp8/block128x128","activation":"bf16","output":"bf16"},"steps":["act bf16","weight fp8/block128x128->bf16","mma bf16*bf16","accumulate f32","scale f32"],"group":235,"kernel":"w8a16_gemm_pipelined_m32::w8a16_gemm_pipelined_m32","family":"w8a16_gemm","point":null,"compute":"tensor_core:mma.sync.m16n8k16.bf16","numerics":"reference","where":"device","state":null},{"id":"l16.gdn.add","order":387,"layer":16,"block":"l16.gdn","block_type":"gdn","kind":"linear_attention","op":"residual_add","roles":[],"inserted":false,"precision":{"weight":null,"activation":"bf16","output":"bf16"},"steps":["compute f32"],"group":236,"kernel":"norm::residual_add_rms_norm","family":"rms_norm","point":null,"compute":"memory","numerics":"reference","where":"device","state":null},{"id":"l16.moe_ffn.post_norm","order":388,"layer":16,"block":"l16.moe_ffn","block_type":"moe_ffn","kind":"linear_attention","op":"rms_norm","roles":[],"inserted":false,"precision":{"weight":null,"activation":"f32","output":"bf16"},"steps":["compute f32"],"group":236,"kernel":"norm::residual_add_rms_norm","family":"rms_norm","point":null,"compute":"memory","numerics":"reference","where":"device","state":null},{"id":"l16.moe_ffn.router","order":389,"layer":16,"block":"l16.moe_ffn","block_type":"moe_ffn","kind":"linear_attention","op":"router","roles":[],"inserted":false,"precision":{"weight":"bf16","activation":"bf16","output":"bf16"},"steps":["act bf16","weight bf16->bf16","mma bf16*bf16","accumulate f32","scale none"],"group":237,"kernel":"gemv::dense_gemv_bf16","family":"dense_bf16","point":null,"compute":"cuda_core","numerics":"reference","where":"device","state":null},{"id":"l16.moe_ffn.top_k","order":390,"layer":16,"block":"l16.moe_ffn","block_type":"moe_ffn","kind":"linear_attention","op":"top_k","roles":[],"inserted":false,"precision":{"weight":null,"activation":"bf16","output":"f32,i32"},"steps":["score f32"],"group":238,"kernel":"moe_topk::moe_topk_softmax","family":"moe_topk","point":null,"compute":"memory","numerics":"reference","where":"device","state":null},{"id":"l16.moe_ffn.experts_gate_up","order":391,"layer":16,"block":"l16.moe_ffn","block_type":"moe_ffn","kind":"linear_attention","op":"expert_gate_up","roles":[],"inserted":false,"precision":{"weight":"fp8/block128x128","activation":"bf16","output":"bf16"},"steps":["gather bf16","act bf16","weight fp8/block128x128->bf16","mma bf16*bf16","accumulate f32","scale f32"],"group":239,"kernel":"moe_shared_expert_fused_fp8::moe_expert_gate_up_shared_fp8","family":"moe_fp8_1row","point":null,"compute":"cuda_core","numerics":"reference","where":"device","state":null},{"id":"l16.moe_ffn.shared_gate_up","order":392,"layer":16,"block":"l16.moe_ffn","block_type":"moe_ffn","kind":"linear_attention","op":"linear:shared_gate_up","roles":["shared_gate_up"],"inserted":false,"precision":{"weight":"fp8/block128x128","activation":"bf16","output":"bf16"},"steps":["act bf16","weight fp8/block128x128->bf16","mma bf16*bf16","accumulate f32","scale f32"],"group":239,"kernel":"moe_shared_expert_fused_fp8::moe_expert_gate_up_shared_fp8","family":"moe_fp8_1row","point":null,"compute":"cuda_core","numerics":"reference","where":"device","state":null},{"id":"l16.moe_ffn.experts_act","order":393,"layer":16,"block":"l16.moe_ffn","block_type":"moe_ffn","kind":"linear_attention","op":"silu_mul","roles":[],"inserted":false,"precision":{"weight":null,"activation":"bf16","output":"f32"},"steps":["compute f32"],"group":240,"kernel":"moe_shared_expert_fused_fp8::moe_expert_silu_down_shared_fp8","family":"moe_fp8_1row","point":null,"compute":"cuda_core","numerics":"reference","where":"device","state":null},{"id":"l16.moe_ffn.experts_down","order":394,"layer":16,"block":"l16.moe_ffn","block_type":"moe_ffn","kind":"linear_attention","op":"expert_down","roles":[],"inserted":false,"precision":{"weight":"fp8/block128x128","activation":"f32","output":"bf16"},"steps":["act f32","weight fp8/block128x128->f32","mma f32*f32","accumulate f32","scale f32"],"group":240,"kernel":"moe_shared_expert_fused_fp8::moe_expert_silu_down_shared_fp8","family":"moe_fp8_1row","point":null,"compute":"cuda_core","numerics":"reference","where":"device","state":null},{"id":"l16.moe_ffn.shared_act","order":395,"layer":16,"block":"l16.moe_ffn","block_type":"moe_ffn","kind":"linear_attention","op":"silu_mul","roles":[],"inserted":false,"precision":{"weight":null,"activation":"bf16","output":"f32"},"steps":["compute f32"],"group":240,"kernel":"moe_shared_expert_fused_fp8::moe_expert_silu_down_shared_fp8","family":"moe_fp8_1row","point":null,"compute":"cuda_core","numerics":"reference","where":"device","state":null},{"id":"l16.moe_ffn.shared_down","order":396,"layer":16,"block":"l16.moe_ffn","block_type":"moe_ffn","kind":"linear_attention","op":"linear:shared_down","roles":["shared_down"],"inserted":false,"precision":{"weight":"fp8/block128x128","activation":"f32","output":"bf16"},"steps":["act f32","weight fp8/block128x128->f32","mma f32*f32","accumulate f32","scale f32"],"group":240,"kernel":"moe_shared_expert_fused_fp8::moe_expert_silu_down_shared_fp8","family":"moe_fp8_1row","point":null,"compute":"cuda_core","numerics":"reference","where":"device","state":null},{"id":"l16.moe_ffn.shared_gate","order":397,"layer":16,"block":"l16.moe_ffn","block_type":"moe_ffn","kind":"linear_attention","op":"linear:shared_gate","roles":["shared_gate"],"inserted":false,"precision":{"weight":"bf16","activation":"bf16","output":"f32"},"steps":["act bf16","weight bf16->bf16","mma bf16*bf16","accumulate f32","scale none"],"group":241,"kernel":"moe_expert_gemv::moe_weighted_sum_blend","family":"moe_blend","point":{"shared_gate":"sigmoid"},"compute":"cuda_core","numerics":"reference","where":"device","state":null},{"id":"l16.moe_ffn.blend","order":398,"layer":16,"block":"l16.moe_ffn","block_type":"moe_ffn","kind":"linear_attention","op":"blend","roles":[],"inserted":false,"precision":{"weight":null,"activation":"bf16","output":"bf16"},"steps":["scatter f32","combine f32"],"group":241,"kernel":"moe_expert_gemv::moe_weighted_sum_blend","family":"moe_blend","point":{"shared_gate":"sigmoid"},"compute":"cuda_core","numerics":"reference","where":"device","state":null},{"id":"l16.moe_ffn.add","order":399,"layer":16,"block":"l16.moe_ffn","block_type":"moe_ffn","kind":"linear_attention","op":"residual_add","roles":[],"inserted":false,"precision":{"weight":null,"activation":"bf16","output":"bf16"},"steps":["compute f32"],"group":242,"kernel":"residual_add_rms_norm_exact::residual_add_rms_norm_exact","family":"rms_norm","point":null,"compute":"memory","numerics":"bit_identical(residual_add_rms_norm_exact_microtest)","where":"device","state":null},{"id":"l17.gdn.input_norm","order":400,"layer":17,"block":"l17.gdn","block_type":"gdn","kind":"linear_attention","op":"rms_norm","roles":[],"inserted":false,"precision":{"weight":null,"activation":"bf16","output":"bf16"},"steps":["compute f32"],"group":242,"kernel":"residual_add_rms_norm_exact::residual_add_rms_norm_exact","family":"rms_norm","point":null,"compute":"memory","numerics":"bit_identical(residual_add_rms_norm_exact_microtest)","where":"device","state":null},{"id":"l17.gdn.qkvz","order":401,"layer":17,"block":"l17.gdn","block_type":"gdn","kind":"linear_attention","op":"linear:qkvz","roles":["qkvz"],"inserted":false,"precision":{"weight":"fp8/block128x128","activation":"bf16","output":"bf16"},"steps":["act bf16","weight fp8/block128x128->bf16","mma bf16*bf16","accumulate f32","scale f32"],"group":243,"kernel":"w8a16_gemm_pipelined_m32::w8a16_gemm_pipelined_m32","family":"w8a16_gemm","point":null,"compute":"tensor_core:mma.sync.m16n8k16.bf16","numerics":"reference","where":"device","state":null},{"id":"l17.gdn.ba","order":402,"layer":17,"block":"l17.gdn","block_type":"gdn","kind":"linear_attention","op":"linear:ba","roles":["ba"],"inserted":false,"precision":{"weight":"bf16","activation":"bf16","output":"f32"},"steps":["act bf16","weight bf16->bf16","mma bf16*bf16","accumulate f32","scale none"],"group":244,"kernel":"ssm_preprocess::dense_gemv_ba_gates","family":"gdn_ba_gates_gemv","point":{"gate":"softplus","decay":"head"},"compute":"cuda_core","numerics":"reference","where":"device","state":null},{"id":"l17.gdn.gates","order":403,"layer":17,"block":"l17.gdn","block_type":"gdn","kind":"linear_attention","op":"gdn_gates","roles":[],"inserted":false,"precision":{"weight":null,"activation":"f32","output":"f32,f32"},"steps":["compute f32"],"group":244,"kernel":"ssm_preprocess::dense_gemv_ba_gates","family":"gdn_ba_gates_gemv","point":{"gate":"softplus","decay":"head"},"compute":"cuda_core","numerics":"reference","where":"device","state":null},{"id":"l17.gdn.conv","order":404,"layer":17,"block":"l17.gdn","block_type":"gdn","kind":"linear_attention","op":"conv1d_update","roles":[],"inserted":false,"precision":{"weight":null,"activation":"bf16","output":"f32"},"steps":["state f32","compute f32"],"group":245,"kernel":"causal_conv1d::causal_conv1d_update_l2norm_f32","family":"causal_conv1d_l2norm","point":null,"compute":"cuda_core","numerics":"reference","where":"device","state":"conv:update"},{"id":"l17.gdn.conv_ckpt","order":405,"layer":17,"block":"l17.gdn","block_type":"gdn","kind":"linear_attention","op":"state_snapshot","roles":[],"inserted":false,"precision":{"weight":null,"activation":"f32","output":null},"steps":["state f32"],"group":245,"kernel":"causal_conv1d::causal_conv1d_update_l2norm_f32","family":"causal_conv1d_l2norm","point":null,"compute":"cuda_core","numerics":"reference","where":"device","state":"conv:snapshot"},{"id":"l17.gdn.l2","order":406,"layer":17,"block":"l17.gdn","block_type":"gdn","kind":"linear_attention","op":"l2_norm","roles":[],"inserted":false,"precision":{"weight":null,"activation":"f32","output":"f32"},"steps":["compute f32"],"group":245,"kernel":"causal_conv1d::causal_conv1d_update_l2norm_f32","family":"causal_conv1d_l2norm","point":null,"compute":"cuda_core","numerics":"reference","where":"device","state":null},{"id":"l17.gdn.recur","order":407,"layer":17,"block":"l17.gdn","block_type":"gdn","kind":"linear_attention","op":"gdn_recurrence","roles":[],"inserted":false,"precision":{"weight":null,"activation":"f32","output":"f32"},"steps":["state f32","compute f32"],"group":246,"kernel":"gated_delta_rule::gated_delta_rule_decode_f32","family":"gdn_recurrence","point":{"decay":"head","gate_act":"silu"},"compute":"cuda_core","numerics":"reference","where":"device","state":"h:update"},{"id":"l17.gdn.out_norm","order":408,"layer":17,"block":"l17.gdn","block_type":"gdn","kind":"linear_attention","op":"gated_rms_norm","roles":[],"inserted":false,"precision":{"weight":null,"activation":"f32","output":"bf16"},"steps":["compute f32"],"group":247,"kernel":"norm::gated_rms_norm_f32_input","family":"gated_rms_norm","point":null,"compute":"memory","numerics":"reference","where":"device","state":null},{"id":"l17.gdn.out","order":409,"layer":17,"block":"l17.gdn","block_type":"gdn","kind":"linear_attention","op":"linear:gdn_out","roles":["gdn_out"],"inserted":false,"precision":{"weight":"fp8/block128x128","activation":"bf16","output":"bf16"},"steps":["act bf16","weight fp8/block128x128->bf16","mma bf16*bf16","accumulate f32","scale f32"],"group":248,"kernel":"w8a16_gemm_pipelined_m32::w8a16_gemm_pipelined_m32","family":"w8a16_gemm","point":null,"compute":"tensor_core:mma.sync.m16n8k16.bf16","numerics":"reference","where":"device","state":null},{"id":"l17.gdn.add","order":410,"layer":17,"block":"l17.gdn","block_type":"gdn","kind":"linear_attention","op":"residual_add","roles":[],"inserted":false,"precision":{"weight":null,"activation":"bf16","output":"bf16"},"steps":["compute f32"],"group":249,"kernel":"norm::residual_add_rms_norm","family":"rms_norm","point":null,"compute":"memory","numerics":"reference","where":"device","state":null},{"id":"l17.moe_ffn.post_norm","order":411,"layer":17,"block":"l17.moe_ffn","block_type":"moe_ffn","kind":"linear_attention","op":"rms_norm","roles":[],"inserted":false,"precision":{"weight":null,"activation":"f32","output":"bf16"},"steps":["compute f32"],"group":249,"kernel":"norm::residual_add_rms_norm","family":"rms_norm","point":null,"compute":"memory","numerics":"reference","where":"device","state":null},{"id":"l17.moe_ffn.router","order":412,"layer":17,"block":"l17.moe_ffn","block_type":"moe_ffn","kind":"linear_attention","op":"router","roles":[],"inserted":false,"precision":{"weight":"bf16","activation":"bf16","output":"bf16"},"steps":["act bf16","weight bf16->bf16","mma bf16*bf16","accumulate f32","scale none"],"group":250,"kernel":"gemv::dense_gemv_bf16","family":"dense_bf16","point":null,"compute":"cuda_core","numerics":"reference","where":"device","state":null},{"id":"l17.moe_ffn.top_k","order":413,"layer":17,"block":"l17.moe_ffn","block_type":"moe_ffn","kind":"linear_attention","op":"top_k","roles":[],"inserted":false,"precision":{"weight":null,"activation":"bf16","output":"f32,i32"},"steps":["score f32"],"group":251,"kernel":"moe_topk::moe_topk_softmax","family":"moe_topk","point":null,"compute":"memory","numerics":"reference","where":"device","state":null},{"id":"l17.moe_ffn.experts_gate_up","order":414,"layer":17,"block":"l17.moe_ffn","block_type":"moe_ffn","kind":"linear_attention","op":"expert_gate_up","roles":[],"inserted":false,"precision":{"weight":"fp8/block128x128","activation":"bf16","output":"bf16"},"steps":["gather bf16","act bf16","weight fp8/block128x128->bf16","mma bf16*bf16","accumulate f32","scale f32"],"group":252,"kernel":"moe_shared_expert_fused_fp8::moe_expert_gate_up_shared_fp8","family":"moe_fp8_1row","point":null,"compute":"cuda_core","numerics":"reference","where":"device","state":null},{"id":"l17.moe_ffn.shared_gate_up","order":415,"layer":17,"block":"l17.moe_ffn","block_type":"moe_ffn","kind":"linear_attention","op":"linear:shared_gate_up","roles":["shared_gate_up"],"inserted":false,"precision":{"weight":"fp8/block128x128","activation":"bf16","output":"bf16"},"steps":["act bf16","weight fp8/block128x128->bf16","mma bf16*bf16","accumulate f32","scale f32"],"group":252,"kernel":"moe_shared_expert_fused_fp8::moe_expert_gate_up_shared_fp8","family":"moe_fp8_1row","point":null,"compute":"cuda_core","numerics":"reference","where":"device","state":null},{"id":"l17.moe_ffn.experts_act","order":416,"layer":17,"block":"l17.moe_ffn","block_type":"moe_ffn","kind":"linear_attention","op":"silu_mul","roles":[],"inserted":false,"precision":{"weight":null,"activation":"bf16","output":"f32"},"steps":["compute f32"],"group":253,"kernel":"moe_shared_expert_fused_fp8::moe_expert_silu_down_shared_fp8","family":"moe_fp8_1row","point":null,"compute":"cuda_core","numerics":"reference","where":"device","state":null},{"id":"l17.moe_ffn.experts_down","order":417,"layer":17,"block":"l17.moe_ffn","block_type":"moe_ffn","kind":"linear_attention","op":"expert_down","roles":[],"inserted":false,"precision":{"weight":"fp8/block128x128","activation":"f32","output":"bf16"},"steps":["act f32","weight fp8/block128x128->f32","mma f32*f32","accumulate f32","scale f32"],"group":253,"kernel":"moe_shared_expert_fused_fp8::moe_expert_silu_down_shared_fp8","family":"moe_fp8_1row","point":null,"compute":"cuda_core","numerics":"reference","where":"device","state":null},{"id":"l17.moe_ffn.shared_act","order":418,"layer":17,"block":"l17.moe_ffn","block_type":"moe_ffn","kind":"linear_attention","op":"silu_mul","roles":[],"inserted":false,"precision":{"weight":null,"activation":"bf16","output":"f32"},"steps":["compute f32"],"group":253,"kernel":"moe_shared_expert_fused_fp8::moe_expert_silu_down_shared_fp8","family":"moe_fp8_1row","point":null,"compute":"cuda_core","numerics":"reference","where":"device","state":null},{"id":"l17.moe_ffn.shared_down","order":419,"layer":17,"block":"l17.moe_ffn","block_type":"moe_ffn","kind":"linear_attention","op":"linear:shared_down","roles":["shared_down"],"inserted":false,"precision":{"weight":"fp8/block128x128","activation":"f32","output":"bf16"},"steps":["act f32","weight fp8/block128x128->f32","mma f32*f32","accumulate f32","scale f32"],"group":253,"kernel":"moe_shared_expert_fused_fp8::moe_expert_silu_down_shared_fp8","family":"moe_fp8_1row","point":null,"compute":"cuda_core","numerics":"reference","where":"device","state":null},{"id":"l17.moe_ffn.shared_gate","order":420,"layer":17,"block":"l17.moe_ffn","block_type":"moe_ffn","kind":"linear_attention","op":"linear:shared_gate","roles":["shared_gate"],"inserted":false,"precision":{"weight":"bf16","activation":"bf16","output":"f32"},"steps":["act bf16","weight bf16->bf16","mma bf16*bf16","accumulate f32","scale none"],"group":254,"kernel":"moe_expert_gemv::moe_weighted_sum_blend","family":"moe_blend","point":{"shared_gate":"sigmoid"},"compute":"cuda_core","numerics":"reference","where":"device","state":null},{"id":"l17.moe_ffn.blend","order":421,"layer":17,"block":"l17.moe_ffn","block_type":"moe_ffn","kind":"linear_attention","op":"blend","roles":[],"inserted":false,"precision":{"weight":null,"activation":"bf16","output":"bf16"},"steps":["scatter f32","combine f32"],"group":254,"kernel":"moe_expert_gemv::moe_weighted_sum_blend","family":"moe_blend","point":{"shared_gate":"sigmoid"},"compute":"cuda_core","numerics":"reference","where":"device","state":null},{"id":"l17.moe_ffn.add","order":422,"layer":17,"block":"l17.moe_ffn","block_type":"moe_ffn","kind":"linear_attention","op":"residual_add","roles":[],"inserted":false,"precision":{"weight":null,"activation":"bf16","output":"bf16"},"steps":["compute f32"],"group":255,"kernel":"residual_add_rms_norm_exact::residual_add_rms_norm_exact","family":"rms_norm","point":null,"compute":"memory","numerics":"bit_identical(residual_add_rms_norm_exact_microtest)","where":"device","state":null},{"id":"l18.gdn.input_norm","order":423,"layer":18,"block":"l18.gdn","block_type":"gdn","kind":"linear_attention","op":"rms_norm","roles":[],"inserted":false,"precision":{"weight":null,"activation":"bf16","output":"bf16"},"steps":["compute f32"],"group":255,"kernel":"residual_add_rms_norm_exact::residual_add_rms_norm_exact","family":"rms_norm","point":null,"compute":"memory","numerics":"bit_identical(residual_add_rms_norm_exact_microtest)","where":"device","state":null},{"id":"l18.gdn.qkvz","order":424,"layer":18,"block":"l18.gdn","block_type":"gdn","kind":"linear_attention","op":"linear:qkvz","roles":["qkvz"],"inserted":false,"precision":{"weight":"fp8/block128x128","activation":"bf16","output":"bf16"},"steps":["act bf16","weight fp8/block128x128->bf16","mma bf16*bf16","accumulate f32","scale f32"],"group":256,"kernel":"w8a16_gemm_pipelined_m32::w8a16_gemm_pipelined_m32","family":"w8a16_gemm","point":null,"compute":"tensor_core:mma.sync.m16n8k16.bf16","numerics":"reference","where":"device","state":null},{"id":"l18.gdn.ba","order":425,"layer":18,"block":"l18.gdn","block_type":"gdn","kind":"linear_attention","op":"linear:ba","roles":["ba"],"inserted":false,"precision":{"weight":"bf16","activation":"bf16","output":"f32"},"steps":["act bf16","weight bf16->bf16","mma bf16*bf16","accumulate f32","scale none"],"group":257,"kernel":"ssm_preprocess::dense_gemv_ba_gates","family":"gdn_ba_gates_gemv","point":{"gate":"softplus","decay":"head"},"compute":"cuda_core","numerics":"reference","where":"device","state":null},{"id":"l18.gdn.gates","order":426,"layer":18,"block":"l18.gdn","block_type":"gdn","kind":"linear_attention","op":"gdn_gates","roles":[],"inserted":false,"precision":{"weight":null,"activation":"f32","output":"f32,f32"},"steps":["compute f32"],"group":257,"kernel":"ssm_preprocess::dense_gemv_ba_gates","family":"gdn_ba_gates_gemv","point":{"gate":"softplus","decay":"head"},"compute":"cuda_core","numerics":"reference","where":"device","state":null},{"id":"l18.gdn.conv","order":427,"layer":18,"block":"l18.gdn","block_type":"gdn","kind":"linear_attention","op":"conv1d_update","roles":[],"inserted":false,"precision":{"weight":null,"activation":"bf16","output":"f32"},"steps":["state f32","compute f32"],"group":258,"kernel":"causal_conv1d::causal_conv1d_update_l2norm_f32","family":"causal_conv1d_l2norm","point":null,"compute":"cuda_core","numerics":"reference","where":"device","state":"conv:update"},{"id":"l18.gdn.conv_ckpt","order":428,"layer":18,"block":"l18.gdn","block_type":"gdn","kind":"linear_attention","op":"state_snapshot","roles":[],"inserted":false,"precision":{"weight":null,"activation":"f32","output":null},"steps":["state f32"],"group":258,"kernel":"causal_conv1d::causal_conv1d_update_l2norm_f32","family":"causal_conv1d_l2norm","point":null,"compute":"cuda_core","numerics":"reference","where":"device","state":"conv:snapshot"},{"id":"l18.gdn.l2","order":429,"layer":18,"block":"l18.gdn","block_type":"gdn","kind":"linear_attention","op":"l2_norm","roles":[],"inserted":false,"precision":{"weight":null,"activation":"f32","output":"f32"},"steps":["compute f32"],"group":258,"kernel":"causal_conv1d::causal_conv1d_update_l2norm_f32","family":"causal_conv1d_l2norm","point":null,"compute":"cuda_core","numerics":"reference","where":"device","state":null},{"id":"l18.gdn.recur","order":430,"layer":18,"block":"l18.gdn","block_type":"gdn","kind":"linear_attention","op":"gdn_recurrence","roles":[],"inserted":false,"precision":{"weight":null,"activation":"f32","output":"f32"},"steps":["state f32","compute f32"],"group":259,"kernel":"gated_delta_rule::gated_delta_rule_decode_f32","family":"gdn_recurrence","point":{"decay":"head","gate_act":"silu"},"compute":"cuda_core","numerics":"reference","where":"device","state":"h:update"},{"id":"l18.gdn.out_norm","order":431,"layer":18,"block":"l18.gdn","block_type":"gdn","kind":"linear_attention","op":"gated_rms_norm","roles":[],"inserted":false,"precision":{"weight":null,"activation":"f32","output":"bf16"},"steps":["compute f32"],"group":260,"kernel":"norm::gated_rms_norm_f32_input","family":"gated_rms_norm","point":null,"compute":"memory","numerics":"reference","where":"device","state":null},{"id":"l18.gdn.out","order":432,"layer":18,"block":"l18.gdn","block_type":"gdn","kind":"linear_attention","op":"linear:gdn_out","roles":["gdn_out"],"inserted":false,"precision":{"weight":"fp8/block128x128","activation":"bf16","output":"bf16"},"steps":["act bf16","weight fp8/block128x128->bf16","mma bf16*bf16","accumulate f32","scale f32"],"group":261,"kernel":"w8a16_gemm_pipelined_m32::w8a16_gemm_pipelined_m32","family":"w8a16_gemm","point":null,"compute":"tensor_core:mma.sync.m16n8k16.bf16","numerics":"reference","where":"device","state":null},{"id":"l18.gdn.add","order":433,"layer":18,"block":"l18.gdn","block_type":"gdn","kind":"linear_attention","op":"residual_add","roles":[],"inserted":false,"precision":{"weight":null,"activation":"bf16","output":"bf16"},"steps":["compute f32"],"group":262,"kernel":"norm::residual_add_rms_norm","family":"rms_norm","point":null,"compute":"memory","numerics":"reference","where":"device","state":null},{"id":"l18.moe_ffn.post_norm","order":434,"layer":18,"block":"l18.moe_ffn","block_type":"moe_ffn","kind":"linear_attention","op":"rms_norm","roles":[],"inserted":false,"precision":{"weight":null,"activation":"f32","output":"bf16"},"steps":["compute f32"],"group":262,"kernel":"norm::residual_add_rms_norm","family":"rms_norm","point":null,"compute":"memory","numerics":"reference","where":"device","state":null},{"id":"l18.moe_ffn.router","order":435,"layer":18,"block":"l18.moe_ffn","block_type":"moe_ffn","kind":"linear_attention","op":"router","roles":[],"inserted":false,"precision":{"weight":"bf16","activation":"bf16","output":"bf16"},"steps":["act bf16","weight bf16->bf16","mma bf16*bf16","accumulate f32","scale none"],"group":263,"kernel":"gemv::dense_gemv_bf16","family":"dense_bf16","point":null,"compute":"cuda_core","numerics":"reference","where":"device","state":null},{"id":"l18.moe_ffn.top_k","order":436,"layer":18,"block":"l18.moe_ffn","block_type":"moe_ffn","kind":"linear_attention","op":"top_k","roles":[],"inserted":false,"precision":{"weight":null,"activation":"bf16","output":"f32,i32"},"steps":["score f32"],"group":264,"kernel":"moe_topk::moe_topk_softmax","family":"moe_topk","point":null,"compute":"memory","numerics":"reference","where":"device","state":null},{"id":"l18.moe_ffn.experts_gate_up","order":437,"layer":18,"block":"l18.moe_ffn","block_type":"moe_ffn","kind":"linear_attention","op":"expert_gate_up","roles":[],"inserted":false,"precision":{"weight":"fp8/block128x128","activation":"bf16","output":"bf16"},"steps":["gather bf16","act bf16","weight fp8/block128x128->bf16","mma bf16*bf16","accumulate f32","scale f32"],"group":265,"kernel":"moe_shared_expert_fused_fp8::moe_expert_gate_up_shared_fp8","family":"moe_fp8_1row","point":null,"compute":"cuda_core","numerics":"reference","where":"device","state":null},{"id":"l18.moe_ffn.shared_gate_up","order":438,"layer":18,"block":"l18.moe_ffn","block_type":"moe_ffn","kind":"linear_attention","op":"linear:shared_gate_up","roles":["shared_gate_up"],"inserted":false,"precision":{"weight":"fp8/block128x128","activation":"bf16","output":"bf16"},"steps":["act bf16","weight fp8/block128x128->bf16","mma bf16*bf16","accumulate f32","scale f32"],"group":265,"kernel":"moe_shared_expert_fused_fp8::moe_expert_gate_up_shared_fp8","family":"moe_fp8_1row","point":null,"compute":"cuda_core","numerics":"reference","where":"device","state":null},{"id":"l18.moe_ffn.experts_act","order":439,"layer":18,"block":"l18.moe_ffn","block_type":"moe_ffn","kind":"linear_attention","op":"silu_mul","roles":[],"inserted":false,"precision":{"weight":null,"activation":"bf16","output":"f32"},"steps":["compute f32"],"group":266,"kernel":"moe_shared_expert_fused_fp8::moe_expert_silu_down_shared_fp8","family":"moe_fp8_1row","point":null,"compute":"cuda_core","numerics":"reference","where":"device","state":null},{"id":"l18.moe_ffn.experts_down","order":440,"layer":18,"block":"l18.moe_ffn","block_type":"moe_ffn","kind":"linear_attention","op":"expert_down","roles":[],"inserted":false,"precision":{"weight":"fp8/block128x128","activation":"f32","output":"bf16"},"steps":["act f32","weight fp8/block128x128->f32","mma f32*f32","accumulate f32","scale f32"],"group":266,"kernel":"moe_shared_expert_fused_fp8::moe_expert_silu_down_shared_fp8","family":"moe_fp8_1row","point":null,"compute":"cuda_core","numerics":"reference","where":"device","state":null},{"id":"l18.moe_ffn.shared_act","order":441,"layer":18,"block":"l18.moe_ffn","block_type":"moe_ffn","kind":"linear_attention","op":"silu_mul","roles":[],"inserted":false,"precision":{"weight":null,"activation":"bf16","output":"f32"},"steps":["compute f32"],"group":266,"kernel":"moe_shared_expert_fused_fp8::moe_expert_silu_down_shared_fp8","family":"moe_fp8_1row","point":null,"compute":"cuda_core","numerics":"reference","where":"device","state":null},{"id":"l18.moe_ffn.shared_down","order":442,"layer":18,"block":"l18.moe_ffn","block_type":"moe_ffn","kind":"linear_attention","op":"linear:shared_down","roles":["shared_down"],"inserted":false,"precision":{"weight":"fp8/block128x128","activation":"f32","output":"bf16"},"steps":["act f32","weight fp8/block128x128->f32","mma f32*f32","accumulate f32","scale f32"],"group":266,"kernel":"moe_shared_expert_fused_fp8::moe_expert_silu_down_shared_fp8","family":"moe_fp8_1row","point":null,"compute":"cuda_core","numerics":"reference","where":"device","state":null},{"id":"l18.moe_ffn.shared_gate","order":443,"layer":18,"block":"l18.moe_ffn","block_type":"moe_ffn","kind":"linear_attention","op":"linear:shared_gate","roles":["shared_gate"],"inserted":false,"precision":{"weight":"bf16","activation":"bf16","output":"f32"},"steps":["act bf16","weight bf16->bf16","mma bf16*bf16","accumulate f32","scale none"],"group":267,"kernel":"moe_expert_gemv::moe_weighted_sum_blend","family":"moe_blend","point":{"shared_gate":"sigmoid"},"compute":"cuda_core","numerics":"reference","where":"device","state":null},{"id":"l18.moe_ffn.blend","order":444,"layer":18,"block":"l18.moe_ffn","block_type":"moe_ffn","kind":"linear_attention","op":"blend","roles":[],"inserted":false,"precision":{"weight":null,"activation":"bf16","output":"bf16"},"steps":["scatter f32","combine f32"],"group":267,"kernel":"moe_expert_gemv::moe_weighted_sum_blend","family":"moe_blend","point":{"shared_gate":"sigmoid"},"compute":"cuda_core","numerics":"reference","where":"device","state":null},{"id":"l18.moe_ffn.add","order":445,"layer":18,"block":"l18.moe_ffn","block_type":"moe_ffn","kind":"linear_attention","op":"residual_add","roles":[],"inserted":false,"precision":{"weight":null,"activation":"bf16","output":"bf16"},"steps":["compute f32"],"group":268,"kernel":"residual_add_rms_norm_exact::residual_add_rms_norm_exact","family":"rms_norm","point":null,"compute":"memory","numerics":"bit_identical(residual_add_rms_norm_exact_microtest)","where":"device","state":null},{"id":"l19.attn.input_norm","order":446,"layer":19,"block":"l19.attn","block_type":"attn","kind":"full_attention","op":"rms_norm","roles":[],"inserted":false,"precision":{"weight":null,"activation":"bf16","output":"bf16"},"steps":["compute f32"],"group":268,"kernel":"residual_add_rms_norm_exact::residual_add_rms_norm_exact","family":"rms_norm","point":null,"compute":"memory","numerics":"bit_identical(residual_add_rms_norm_exact_microtest)","where":"device","state":null},{"id":"l19.attn.q","order":447,"layer":19,"block":"l19.attn","block_type":"attn","kind":"full_attention","op":"linear:q","roles":["q"],"inserted":false,"precision":{"weight":"fp8/block128x128","activation":"bf16","output":"bf16"},"steps":["act bf16","weight fp8/block128x128->bf16","mma bf16*bf16","accumulate f32","scale f32"],"group":269,"kernel":"w8a16_gemm_pipelined_m32::w8a16_gemm_pipelined_m32","family":"w8a16_gemm","point":null,"compute":"tensor_core:mma.sync.m16n8k16.bf16","numerics":"reference","where":"device","state":null},{"id":"l19.attn.q_split","order":448,"layer":19,"block":"l19.attn","block_type":"attn","kind":"full_attention","op":"split","roles":[],"inserted":false,"precision":{"weight":null,"activation":"bf16","output":"bf16,bf16"},"steps":["move bf16"],"group":270,"kernel":"ssm_preprocess::deinterleave_qg","family":"deinterleave_qg","point":{"layout":"interleaved"},"compute":"memory","numerics":"reference","where":"device","state":null},{"id":"l19.attn.k","order":449,"layer":19,"block":"l19.attn","block_type":"attn","kind":"full_attention","op":"linear:k","roles":["k"],"inserted":false,"precision":{"weight":"fp8/block128x128","activation":"bf16","output":"bf16"},"steps":["act bf16","weight fp8/block128x128->bf16","mma bf16*bf16","accumulate f32","scale f32"],"group":271,"kernel":"w8a16_gemm_pipelined_m32::w8a16_gemm_pipelined_m32","family":"w8a16_gemm","point":null,"compute":"tensor_core:mma.sync.m16n8k16.bf16","numerics":"reference","where":"device","state":null},{"id":"l19.attn.v","order":450,"layer":19,"block":"l19.attn","block_type":"attn","kind":"full_attention","op":"linear:v","roles":["v"],"inserted":false,"precision":{"weight":"fp8/block128x128","activation":"bf16","output":"bf16"},"steps":["act bf16","weight fp8/block128x128->bf16","mma bf16*bf16","accumulate f32","scale f32"],"group":272,"kernel":"w8a16_gemm_pipelined_m32::w8a16_gemm_pipelined_m32","family":"w8a16_gemm","point":null,"compute":"tensor_core:mma.sync.m16n8k16.bf16","numerics":"reference","where":"device","state":null},{"id":"l19.attn.q_norm","order":451,"layer":19,"block":"l19.attn","block_type":"attn","kind":"full_attention","op":"qk_norm","roles":[],"inserted":false,"precision":{"weight":null,"activation":"bf16","output":"bf16"},"steps":["compute f32"],"group":273,"kernel":"norm::rms_norm","family":"rms_norm","point":null,"compute":"memory","numerics":"reference","where":"device","state":null},{"id":"l19.attn.k_norm","order":452,"layer":19,"block":"l19.attn","block_type":"attn","kind":"full_attention","op":"qk_norm","roles":[],"inserted":false,"precision":{"weight":null,"activation":"bf16","output":"bf16"},"steps":["compute f32"],"group":274,"kernel":"norm::rms_norm","family":"rms_norm","point":null,"compute":"memory","numerics":"reference","where":"device","state":null},{"id":"l19.attn.rope","order":453,"layer":19,"block":"l19.attn","block_type":"attn","kind":"full_attention","op":"rope","roles":[],"inserted":false,"precision":{"weight":null,"activation":"bf16","output":"bf16,bf16"},"steps":["compute f32"],"group":275,"kernel":"rope_mrope_interleaved::rope_forward_mrope_interleaved","family":"rope","point":null,"compute":"memory","numerics":"reference","where":"device","state":null},{"id":"l19.attn.kv_write","order":454,"layer":19,"block":"l19.attn","block_type":"attn","kind":"full_attention","op":"kv_write","roles":[],"inserted":false,"precision":{"weight":null,"activation":"bf16","output":null},"steps":["cache bf16"],"group":276,"kernel":"reshape_and_cache::reshape_and_cache_flash","family":"kv_write","point":null,"compute":"memory","numerics":"reference","where":"device","state":"k:write,v:write"},{"id":"l19.attn.attend","order":455,"layer":19,"block":"l19.attn","block_type":"attn","kind":"full_attention","op":"paged_attention","roles":[],"inserted":false,"precision":{"weight":null,"activation":"bf16","output":"bf16"},"steps":["cache bf16","scores f32","softmax f32","accumulate f32"],"group":277,"kernel":"paged_decode::paged_decode_attn","family":"paged_decode_attn","point":null,"compute":"cuda_core","numerics":"reference","where":"device","state":"k:read,v:read"},{"id":"l19.attn.gate_mul","order":456,"layer":19,"block":"l19.attn","block_type":"attn","kind":"full_attention","op":"sigmoid_gate_mul","roles":[],"inserted":false,"precision":{"weight":null,"activation":"bf16","output":"bf16"},"steps":["compute f32"],"group":278,"kernel":"residual_add::sigmoid_gate_mul","family":"sigmoid_gate_mul","point":null,"compute":"memory","numerics":"reference","where":"device","state":null},{"id":"l19.attn.o","order":457,"layer":19,"block":"l19.attn","block_type":"attn","kind":"full_attention","op":"linear:o","roles":["o"],"inserted":false,"precision":{"weight":"fp8/block128x128","activation":"bf16","output":"bf16"},"steps":["act bf16","weight fp8/block128x128->bf16","mma bf16*bf16","accumulate f32","scale f32"],"group":279,"kernel":"w8a16_gemm_pipelined_m32::w8a16_gemm_pipelined_m32","family":"w8a16_gemm","point":null,"compute":"tensor_core:mma.sync.m16n8k16.bf16","numerics":"reference","where":"device","state":null},{"id":"l19.attn.add","order":458,"layer":19,"block":"l19.attn","block_type":"attn","kind":"full_attention","op":"residual_add","roles":[],"inserted":false,"precision":{"weight":null,"activation":"bf16","output":"bf16"},"steps":["compute f32"],"group":280,"kernel":"norm::residual_add_rms_norm","family":"rms_norm","point":null,"compute":"memory","numerics":"reference","where":"device","state":null},{"id":"l19.moe_ffn.post_norm","order":459,"layer":19,"block":"l19.moe_ffn","block_type":"moe_ffn","kind":"full_attention","op":"rms_norm","roles":[],"inserted":false,"precision":{"weight":null,"activation":"f32","output":"bf16"},"steps":["compute f32"],"group":280,"kernel":"norm::residual_add_rms_norm","family":"rms_norm","point":null,"compute":"memory","numerics":"reference","where":"device","state":null},{"id":"l19.moe_ffn.router","order":460,"layer":19,"block":"l19.moe_ffn","block_type":"moe_ffn","kind":"full_attention","op":"router","roles":[],"inserted":false,"precision":{"weight":"bf16","activation":"bf16","output":"bf16"},"steps":["act bf16","weight bf16->bf16","mma bf16*bf16","accumulate f32","scale none"],"group":281,"kernel":"gemv::dense_gemv_bf16","family":"dense_bf16","point":null,"compute":"cuda_core","numerics":"reference","where":"device","state":null},{"id":"l19.moe_ffn.top_k","order":461,"layer":19,"block":"l19.moe_ffn","block_type":"moe_ffn","kind":"full_attention","op":"top_k","roles":[],"inserted":false,"precision":{"weight":null,"activation":"bf16","output":"f32,i32"},"steps":["score f32"],"group":282,"kernel":"moe_topk::moe_topk_softmax","family":"moe_topk","point":null,"compute":"memory","numerics":"reference","where":"device","state":null},{"id":"l19.moe_ffn.experts_gate_up","order":462,"layer":19,"block":"l19.moe_ffn","block_type":"moe_ffn","kind":"full_attention","op":"expert_gate_up","roles":[],"inserted":false,"precision":{"weight":"fp8/block128x128","activation":"bf16","output":"bf16"},"steps":["gather bf16","act bf16","weight fp8/block128x128->bf16","mma bf16*bf16","accumulate f32","scale f32"],"group":283,"kernel":"moe_shared_expert_fused_fp8::moe_expert_gate_up_shared_fp8","family":"moe_fp8_1row","point":null,"compute":"cuda_core","numerics":"reference","where":"device","state":null},{"id":"l19.moe_ffn.shared_gate_up","order":463,"layer":19,"block":"l19.moe_ffn","block_type":"moe_ffn","kind":"full_attention","op":"linear:shared_gate_up","roles":["shared_gate_up"],"inserted":false,"precision":{"weight":"fp8/block128x128","activation":"bf16","output":"bf16"},"steps":["act bf16","weight fp8/block128x128->bf16","mma bf16*bf16","accumulate f32","scale f32"],"group":283,"kernel":"moe_shared_expert_fused_fp8::moe_expert_gate_up_shared_fp8","family":"moe_fp8_1row","point":null,"compute":"cuda_core","numerics":"reference","where":"device","state":null},{"id":"l19.moe_ffn.experts_act","order":464,"layer":19,"block":"l19.moe_ffn","block_type":"moe_ffn","kind":"full_attention","op":"silu_mul","roles":[],"inserted":false,"precision":{"weight":null,"activation":"bf16","output":"f32"},"steps":["compute f32"],"group":284,"kernel":"moe_shared_expert_fused_fp8::moe_expert_silu_down_shared_fp8","family":"moe_fp8_1row","point":null,"compute":"cuda_core","numerics":"reference","where":"device","state":null},{"id":"l19.moe_ffn.experts_down","order":465,"layer":19,"block":"l19.moe_ffn","block_type":"moe_ffn","kind":"full_attention","op":"expert_down","roles":[],"inserted":false,"precision":{"weight":"fp8/block128x128","activation":"f32","output":"bf16"},"steps":["act f32","weight fp8/block128x128->f32","mma f32*f32","accumulate f32","scale f32"],"group":284,"kernel":"moe_shared_expert_fused_fp8::moe_expert_silu_down_shared_fp8","family":"moe_fp8_1row","point":null,"compute":"cuda_core","numerics":"reference","where":"device","state":null},{"id":"l19.moe_ffn.shared_act","order":466,"layer":19,"block":"l19.moe_ffn","block_type":"moe_ffn","kind":"full_attention","op":"silu_mul","roles":[],"inserted":false,"precision":{"weight":null,"activation":"bf16","output":"f32"},"steps":["compute f32"],"group":284,"kernel":"moe_shared_expert_fused_fp8::moe_expert_silu_down_shared_fp8","family":"moe_fp8_1row","point":null,"compute":"cuda_core","numerics":"reference","where":"device","state":null},{"id":"l19.moe_ffn.shared_down","order":467,"layer":19,"block":"l19.moe_ffn","block_type":"moe_ffn","kind":"full_attention","op":"linear:shared_down","roles":["shared_down"],"inserted":false,"precision":{"weight":"fp8/block128x128","activation":"f32","output":"bf16"},"steps":["act f32","weight fp8/block128x128->f32","mma f32*f32","accumulate f32","scale f32"],"group":284,"kernel":"moe_shared_expert_fused_fp8::moe_expert_silu_down_shared_fp8","family":"moe_fp8_1row","point":null,"compute":"cuda_core","numerics":"reference","where":"device","state":null},{"id":"l19.moe_ffn.shared_gate","order":468,"layer":19,"block":"l19.moe_ffn","block_type":"moe_ffn","kind":"full_attention","op":"linear:shared_gate","roles":["shared_gate"],"inserted":false,"precision":{"weight":"bf16","activation":"bf16","output":"f32"},"steps":["act bf16","weight bf16->bf16","mma bf16*bf16","accumulate f32","scale none"],"group":285,"kernel":"moe_expert_gemv::moe_weighted_sum_blend","family":"moe_blend","point":{"shared_gate":"sigmoid"},"compute":"cuda_core","numerics":"reference","where":"device","state":null},{"id":"l19.moe_ffn.blend","order":469,"layer":19,"block":"l19.moe_ffn","block_type":"moe_ffn","kind":"full_attention","op":"blend","roles":[],"inserted":false,"precision":{"weight":null,"activation":"bf16","output":"bf16"},"steps":["scatter f32","combine f32"],"group":285,"kernel":"moe_expert_gemv::moe_weighted_sum_blend","family":"moe_blend","point":{"shared_gate":"sigmoid"},"compute":"cuda_core","numerics":"reference","where":"device","state":null},{"id":"l19.moe_ffn.add","order":470,"layer":19,"block":"l19.moe_ffn","block_type":"moe_ffn","kind":"full_attention","op":"residual_add","roles":[],"inserted":false,"precision":{"weight":null,"activation":"bf16","output":"bf16"},"steps":["compute f32"],"group":286,"kernel":"residual_add_rms_norm_exact::residual_add_rms_norm_exact","family":"rms_norm","point":null,"compute":"memory","numerics":"bit_identical(residual_add_rms_norm_exact_microtest)","where":"device","state":null},{"id":"l20.gdn.input_norm","order":471,"layer":20,"block":"l20.gdn","block_type":"gdn","kind":"linear_attention","op":"rms_norm","roles":[],"inserted":false,"precision":{"weight":null,"activation":"bf16","output":"bf16"},"steps":["compute f32"],"group":286,"kernel":"residual_add_rms_norm_exact::residual_add_rms_norm_exact","family":"rms_norm","point":null,"compute":"memory","numerics":"bit_identical(residual_add_rms_norm_exact_microtest)","where":"device","state":null},{"id":"l20.gdn.qkvz","order":472,"layer":20,"block":"l20.gdn","block_type":"gdn","kind":"linear_attention","op":"linear:qkvz","roles":["qkvz"],"inserted":false,"precision":{"weight":"fp8/block128x128","activation":"bf16","output":"bf16"},"steps":["act bf16","weight fp8/block128x128->bf16","mma bf16*bf16","accumulate f32","scale f32"],"group":287,"kernel":"w8a16_gemm_pipelined_m32::w8a16_gemm_pipelined_m32","family":"w8a16_gemm","point":null,"compute":"tensor_core:mma.sync.m16n8k16.bf16","numerics":"reference","where":"device","state":null},{"id":"l20.gdn.ba","order":473,"layer":20,"block":"l20.gdn","block_type":"gdn","kind":"linear_attention","op":"linear:ba","roles":["ba"],"inserted":false,"precision":{"weight":"bf16","activation":"bf16","output":"f32"},"steps":["act bf16","weight bf16->bf16","mma bf16*bf16","accumulate f32","scale none"],"group":288,"kernel":"ssm_preprocess::dense_gemv_ba_gates","family":"gdn_ba_gates_gemv","point":{"gate":"softplus","decay":"head"},"compute":"cuda_core","numerics":"reference","where":"device","state":null},{"id":"l20.gdn.gates","order":474,"layer":20,"block":"l20.gdn","block_type":"gdn","kind":"linear_attention","op":"gdn_gates","roles":[],"inserted":false,"precision":{"weight":null,"activation":"f32","output":"f32,f32"},"steps":["compute f32"],"group":288,"kernel":"ssm_preprocess::dense_gemv_ba_gates","family":"gdn_ba_gates_gemv","point":{"gate":"softplus","decay":"head"},"compute":"cuda_core","numerics":"reference","where":"device","state":null},{"id":"l20.gdn.conv","order":475,"layer":20,"block":"l20.gdn","block_type":"gdn","kind":"linear_attention","op":"conv1d_update","roles":[],"inserted":false,"precision":{"weight":null,"activation":"bf16","output":"f32"},"steps":["state f32","compute f32"],"group":289,"kernel":"causal_conv1d::causal_conv1d_update_l2norm_f32","family":"causal_conv1d_l2norm","point":null,"compute":"cuda_core","numerics":"reference","where":"device","state":"conv:update"},{"id":"l20.gdn.conv_ckpt","order":476,"layer":20,"block":"l20.gdn","block_type":"gdn","kind":"linear_attention","op":"state_snapshot","roles":[],"inserted":false,"precision":{"weight":null,"activation":"f32","output":null},"steps":["state f32"],"group":289,"kernel":"causal_conv1d::causal_conv1d_update_l2norm_f32","family":"causal_conv1d_l2norm","point":null,"compute":"cuda_core","numerics":"reference","where":"device","state":"conv:snapshot"},{"id":"l20.gdn.l2","order":477,"layer":20,"block":"l20.gdn","block_type":"gdn","kind":"linear_attention","op":"l2_norm","roles":[],"inserted":false,"precision":{"weight":null,"activation":"f32","output":"f32"},"steps":["compute f32"],"group":289,"kernel":"causal_conv1d::causal_conv1d_update_l2norm_f32","family":"causal_conv1d_l2norm","point":null,"compute":"cuda_core","numerics":"reference","where":"device","state":null},{"id":"l20.gdn.recur","order":478,"layer":20,"block":"l20.gdn","block_type":"gdn","kind":"linear_attention","op":"gdn_recurrence","roles":[],"inserted":false,"precision":{"weight":null,"activation":"f32","output":"f32"},"steps":["state f32","compute f32"],"group":290,"kernel":"gated_delta_rule::gated_delta_rule_decode_f32","family":"gdn_recurrence","point":{"decay":"head","gate_act":"silu"},"compute":"cuda_core","numerics":"reference","where":"device","state":"h:update"},{"id":"l20.gdn.out_norm","order":479,"layer":20,"block":"l20.gdn","block_type":"gdn","kind":"linear_attention","op":"gated_rms_norm","roles":[],"inserted":false,"precision":{"weight":null,"activation":"f32","output":"bf16"},"steps":["compute f32"],"group":291,"kernel":"norm::gated_rms_norm_f32_input","family":"gated_rms_norm","point":null,"compute":"memory","numerics":"reference","where":"device","state":null},{"id":"l20.gdn.out","order":480,"layer":20,"block":"l20.gdn","block_type":"gdn","kind":"linear_attention","op":"linear:gdn_out","roles":["gdn_out"],"inserted":false,"precision":{"weight":"fp8/block128x128","activation":"bf16","output":"bf16"},"steps":["act bf16","weight fp8/block128x128->bf16","mma bf16*bf16","accumulate f32","scale f32"],"group":292,"kernel":"w8a16_gemm_pipelined_m32::w8a16_gemm_pipelined_m32","family":"w8a16_gemm","point":null,"compute":"tensor_core:mma.sync.m16n8k16.bf16","numerics":"reference","where":"device","state":null},{"id":"l20.gdn.add","order":481,"layer":20,"block":"l20.gdn","block_type":"gdn","kind":"linear_attention","op":"residual_add","roles":[],"inserted":false,"precision":{"weight":null,"activation":"bf16","output":"bf16"},"steps":["compute f32"],"group":293,"kernel":"norm::residual_add_rms_norm","family":"rms_norm","point":null,"compute":"memory","numerics":"reference","where":"device","state":null},{"id":"l20.moe_ffn.post_norm","order":482,"layer":20,"block":"l20.moe_ffn","block_type":"moe_ffn","kind":"linear_attention","op":"rms_norm","roles":[],"inserted":false,"precision":{"weight":null,"activation":"f32","output":"bf16"},"steps":["compute f32"],"group":293,"kernel":"norm::residual_add_rms_norm","family":"rms_norm","point":null,"compute":"memory","numerics":"reference","where":"device","state":null},{"id":"l20.moe_ffn.router","order":483,"layer":20,"block":"l20.moe_ffn","block_type":"moe_ffn","kind":"linear_attention","op":"router","roles":[],"inserted":false,"precision":{"weight":"bf16","activation":"bf16","output":"bf16"},"steps":["act bf16","weight bf16->bf16","mma bf16*bf16","accumulate f32","scale none"],"group":294,"kernel":"gemv::dense_gemv_bf16","family":"dense_bf16","point":null,"compute":"cuda_core","numerics":"reference","where":"device","state":null},{"id":"l20.moe_ffn.top_k","order":484,"layer":20,"block":"l20.moe_ffn","block_type":"moe_ffn","kind":"linear_attention","op":"top_k","roles":[],"inserted":false,"precision":{"weight":null,"activation":"bf16","output":"f32,i32"},"steps":["score f32"],"group":295,"kernel":"moe_topk::moe_topk_softmax","family":"moe_topk","point":null,"compute":"memory","numerics":"reference","where":"device","state":null},{"id":"l20.moe_ffn.experts_gate_up","order":485,"layer":20,"block":"l20.moe_ffn","block_type":"moe_ffn","kind":"linear_attention","op":"expert_gate_up","roles":[],"inserted":false,"precision":{"weight":"fp8/block128x128","activation":"bf16","output":"bf16"},"steps":["gather bf16","act bf16","weight fp8/block128x128->bf16","mma bf16*bf16","accumulate f32","scale f32"],"group":296,"kernel":"moe_shared_expert_fused_fp8::moe_expert_gate_up_shared_fp8","family":"moe_fp8_1row","point":null,"compute":"cuda_core","numerics":"reference","where":"device","state":null},{"id":"l20.moe_ffn.shared_gate_up","order":486,"layer":20,"block":"l20.moe_ffn","block_type":"moe_ffn","kind":"linear_attention","op":"linear:shared_gate_up","roles":["shared_gate_up"],"inserted":false,"precision":{"weight":"fp8/block128x128","activation":"bf16","output":"bf16"},"steps":["act bf16","weight fp8/block128x128->bf16","mma bf16*bf16","accumulate f32","scale f32"],"group":296,"kernel":"moe_shared_expert_fused_fp8::moe_expert_gate_up_shared_fp8","family":"moe_fp8_1row","point":null,"compute":"cuda_core","numerics":"reference","where":"device","state":null},{"id":"l20.moe_ffn.experts_act","order":487,"layer":20,"block":"l20.moe_ffn","block_type":"moe_ffn","kind":"linear_attention","op":"silu_mul","roles":[],"inserted":false,"precision":{"weight":null,"activation":"bf16","output":"f32"},"steps":["compute f32"],"group":297,"kernel":"moe_shared_expert_fused_fp8::moe_expert_silu_down_shared_fp8","family":"moe_fp8_1row","point":null,"compute":"cuda_core","numerics":"reference","where":"device","state":null},{"id":"l20.moe_ffn.experts_down","order":488,"layer":20,"block":"l20.moe_ffn","block_type":"moe_ffn","kind":"linear_attention","op":"expert_down","roles":[],"inserted":false,"precision":{"weight":"fp8/block128x128","activation":"f32","output":"bf16"},"steps":["act f32","weight fp8/block128x128->f32","mma f32*f32","accumulate f32","scale f32"],"group":297,"kernel":"moe_shared_expert_fused_fp8::moe_expert_silu_down_shared_fp8","family":"moe_fp8_1row","point":null,"compute":"cuda_core","numerics":"reference","where":"device","state":null},{"id":"l20.moe_ffn.shared_act","order":489,"layer":20,"block":"l20.moe_ffn","block_type":"moe_ffn","kind":"linear_attention","op":"silu_mul","roles":[],"inserted":false,"precision":{"weight":null,"activation":"bf16","output":"f32"},"steps":["compute f32"],"group":297,"kernel":"moe_shared_expert_fused_fp8::moe_expert_silu_down_shared_fp8","family":"moe_fp8_1row","point":null,"compute":"cuda_core","numerics":"reference","where":"device","state":null},{"id":"l20.moe_ffn.shared_down","order":490,"layer":20,"block":"l20.moe_ffn","block_type":"moe_ffn","kind":"linear_attention","op":"linear:shared_down","roles":["shared_down"],"inserted":false,"precision":{"weight":"fp8/block128x128","activation":"f32","output":"bf16"},"steps":["act f32","weight fp8/block128x128->f32","mma f32*f32","accumulate f32","scale f32"],"group":297,"kernel":"moe_shared_expert_fused_fp8::moe_expert_silu_down_shared_fp8","family":"moe_fp8_1row","point":null,"compute":"cuda_core","numerics":"reference","where":"device","state":null},{"id":"l20.moe_ffn.shared_gate","order":491,"layer":20,"block":"l20.moe_ffn","block_type":"moe_ffn","kind":"linear_attention","op":"linear:shared_gate","roles":["shared_gate"],"inserted":false,"precision":{"weight":"bf16","activation":"bf16","output":"f32"},"steps":["act bf16","weight bf16->bf16","mma bf16*bf16","accumulate f32","scale none"],"group":298,"kernel":"moe_expert_gemv::moe_weighted_sum_blend","family":"moe_blend","point":{"shared_gate":"sigmoid"},"compute":"cuda_core","numerics":"reference","where":"device","state":null},{"id":"l20.moe_ffn.blend","order":492,"layer":20,"block":"l20.moe_ffn","block_type":"moe_ffn","kind":"linear_attention","op":"blend","roles":[],"inserted":false,"precision":{"weight":null,"activation":"bf16","output":"bf16"},"steps":["scatter f32","combine f32"],"group":298,"kernel":"moe_expert_gemv::moe_weighted_sum_blend","family":"moe_blend","point":{"shared_gate":"sigmoid"},"compute":"cuda_core","numerics":"reference","where":"device","state":null},{"id":"l20.moe_ffn.add","order":493,"layer":20,"block":"l20.moe_ffn","block_type":"moe_ffn","kind":"linear_attention","op":"residual_add","roles":[],"inserted":false,"precision":{"weight":null,"activation":"bf16","output":"bf16"},"steps":["compute f32"],"group":299,"kernel":"residual_add_rms_norm_exact::residual_add_rms_norm_exact","family":"rms_norm","point":null,"compute":"memory","numerics":"bit_identical(residual_add_rms_norm_exact_microtest)","where":"device","state":null},{"id":"l21.gdn.input_norm","order":494,"layer":21,"block":"l21.gdn","block_type":"gdn","kind":"linear_attention","op":"rms_norm","roles":[],"inserted":false,"precision":{"weight":null,"activation":"bf16","output":"bf16"},"steps":["compute f32"],"group":299,"kernel":"residual_add_rms_norm_exact::residual_add_rms_norm_exact","family":"rms_norm","point":null,"compute":"memory","numerics":"bit_identical(residual_add_rms_norm_exact_microtest)","where":"device","state":null},{"id":"l21.gdn.qkvz","order":495,"layer":21,"block":"l21.gdn","block_type":"gdn","kind":"linear_attention","op":"linear:qkvz","roles":["qkvz"],"inserted":false,"precision":{"weight":"fp8/block128x128","activation":"bf16","output":"bf16"},"steps":["act bf16","weight fp8/block128x128->bf16","mma bf16*bf16","accumulate f32","scale f32"],"group":300,"kernel":"w8a16_gemm_pipelined_m32::w8a16_gemm_pipelined_m32","family":"w8a16_gemm","point":null,"compute":"tensor_core:mma.sync.m16n8k16.bf16","numerics":"reference","where":"device","state":null},{"id":"l21.gdn.ba","order":496,"layer":21,"block":"l21.gdn","block_type":"gdn","kind":"linear_attention","op":"linear:ba","roles":["ba"],"inserted":false,"precision":{"weight":"bf16","activation":"bf16","output":"f32"},"steps":["act bf16","weight bf16->bf16","mma bf16*bf16","accumulate f32","scale none"],"group":301,"kernel":"ssm_preprocess::dense_gemv_ba_gates","family":"gdn_ba_gates_gemv","point":{"gate":"softplus","decay":"head"},"compute":"cuda_core","numerics":"reference","where":"device","state":null},{"id":"l21.gdn.gates","order":497,"layer":21,"block":"l21.gdn","block_type":"gdn","kind":"linear_attention","op":"gdn_gates","roles":[],"inserted":false,"precision":{"weight":null,"activation":"f32","output":"f32,f32"},"steps":["compute f32"],"group":301,"kernel":"ssm_preprocess::dense_gemv_ba_gates","family":"gdn_ba_gates_gemv","point":{"gate":"softplus","decay":"head"},"compute":"cuda_core","numerics":"reference","where":"device","state":null},{"id":"l21.gdn.conv","order":498,"layer":21,"block":"l21.gdn","block_type":"gdn","kind":"linear_attention","op":"conv1d_update","roles":[],"inserted":false,"precision":{"weight":null,"activation":"bf16","output":"f32"},"steps":["state f32","compute f32"],"group":302,"kernel":"causal_conv1d::causal_conv1d_update_l2norm_f32","family":"causal_conv1d_l2norm","point":null,"compute":"cuda_core","numerics":"reference","where":"device","state":"conv:update"},{"id":"l21.gdn.conv_ckpt","order":499,"layer":21,"block":"l21.gdn","block_type":"gdn","kind":"linear_attention","op":"state_snapshot","roles":[],"inserted":false,"precision":{"weight":null,"activation":"f32","output":null},"steps":["state f32"],"group":302,"kernel":"causal_conv1d::causal_conv1d_update_l2norm_f32","family":"causal_conv1d_l2norm","point":null,"compute":"cuda_core","numerics":"reference","where":"device","state":"conv:snapshot"},{"id":"l21.gdn.l2","order":500,"layer":21,"block":"l21.gdn","block_type":"gdn","kind":"linear_attention","op":"l2_norm","roles":[],"inserted":false,"precision":{"weight":null,"activation":"f32","output":"f32"},"steps":["compute f32"],"group":302,"kernel":"causal_conv1d::causal_conv1d_update_l2norm_f32","family":"causal_conv1d_l2norm","point":null,"compute":"cuda_core","numerics":"reference","where":"device","state":null},{"id":"l21.gdn.recur","order":501,"layer":21,"block":"l21.gdn","block_type":"gdn","kind":"linear_attention","op":"gdn_recurrence","roles":[],"inserted":false,"precision":{"weight":null,"activation":"f32","output":"f32"},"steps":["state f32","compute f32"],"group":303,"kernel":"gated_delta_rule::gated_delta_rule_decode_f32","family":"gdn_recurrence","point":{"decay":"head","gate_act":"silu"},"compute":"cuda_core","numerics":"reference","where":"device","state":"h:update"},{"id":"l21.gdn.out_norm","order":502,"layer":21,"block":"l21.gdn","block_type":"gdn","kind":"linear_attention","op":"gated_rms_norm","roles":[],"inserted":false,"precision":{"weight":null,"activation":"f32","output":"bf16"},"steps":["compute f32"],"group":304,"kernel":"norm::gated_rms_norm_f32_input","family":"gated_rms_norm","point":null,"compute":"memory","numerics":"reference","where":"device","state":null},{"id":"l21.gdn.out","order":503,"layer":21,"block":"l21.gdn","block_type":"gdn","kind":"linear_attention","op":"linear:gdn_out","roles":["gdn_out"],"inserted":false,"precision":{"weight":"fp8/block128x128","activation":"bf16","output":"bf16"},"steps":["act bf16","weight fp8/block128x128->bf16","mma bf16*bf16","accumulate f32","scale f32"],"group":305,"kernel":"w8a16_gemm_pipelined_m32::w8a16_gemm_pipelined_m32","family":"w8a16_gemm","point":null,"compute":"tensor_core:mma.sync.m16n8k16.bf16","numerics":"reference","where":"device","state":null},{"id":"l21.gdn.add","order":504,"layer":21,"block":"l21.gdn","block_type":"gdn","kind":"linear_attention","op":"residual_add","roles":[],"inserted":false,"precision":{"weight":null,"activation":"bf16","output":"bf16"},"steps":["compute f32"],"group":306,"kernel":"norm::residual_add_rms_norm","family":"rms_norm","point":null,"compute":"memory","numerics":"reference","where":"device","state":null},{"id":"l21.moe_ffn.post_norm","order":505,"layer":21,"block":"l21.moe_ffn","block_type":"moe_ffn","kind":"linear_attention","op":"rms_norm","roles":[],"inserted":false,"precision":{"weight":null,"activation":"f32","output":"bf16"},"steps":["compute f32"],"group":306,"kernel":"norm::residual_add_rms_norm","family":"rms_norm","point":null,"compute":"memory","numerics":"reference","where":"device","state":null},{"id":"l21.moe_ffn.router","order":506,"layer":21,"block":"l21.moe_ffn","block_type":"moe_ffn","kind":"linear_attention","op":"router","roles":[],"inserted":false,"precision":{"weight":"bf16","activation":"bf16","output":"bf16"},"steps":["act bf16","weight bf16->bf16","mma bf16*bf16","accumulate f32","scale none"],"group":307,"kernel":"gemv::dense_gemv_bf16","family":"dense_bf16","point":null,"compute":"cuda_core","numerics":"reference","where":"device","state":null},{"id":"l21.moe_ffn.top_k","order":507,"layer":21,"block":"l21.moe_ffn","block_type":"moe_ffn","kind":"linear_attention","op":"top_k","roles":[],"inserted":false,"precision":{"weight":null,"activation":"bf16","output":"f32,i32"},"steps":["score f32"],"group":308,"kernel":"moe_topk::moe_topk_softmax","family":"moe_topk","point":null,"compute":"memory","numerics":"reference","where":"device","state":null},{"id":"l21.moe_ffn.experts_gate_up","order":508,"layer":21,"block":"l21.moe_ffn","block_type":"moe_ffn","kind":"linear_attention","op":"expert_gate_up","roles":[],"inserted":false,"precision":{"weight":"fp8/block128x128","activation":"bf16","output":"bf16"},"steps":["gather bf16","act bf16","weight fp8/block128x128->bf16","mma bf16*bf16","accumulate f32","scale f32"],"group":309,"kernel":"moe_shared_expert_fused_fp8::moe_expert_gate_up_shared_fp8","family":"moe_fp8_1row","point":null,"compute":"cuda_core","numerics":"reference","where":"device","state":null},{"id":"l21.moe_ffn.shared_gate_up","order":509,"layer":21,"block":"l21.moe_ffn","block_type":"moe_ffn","kind":"linear_attention","op":"linear:shared_gate_up","roles":["shared_gate_up"],"inserted":false,"precision":{"weight":"fp8/block128x128","activation":"bf16","output":"bf16"},"steps":["act bf16","weight fp8/block128x128->bf16","mma bf16*bf16","accumulate f32","scale f32"],"group":309,"kernel":"moe_shared_expert_fused_fp8::moe_expert_gate_up_shared_fp8","family":"moe_fp8_1row","point":null,"compute":"cuda_core","numerics":"reference","where":"device","state":null},{"id":"l21.moe_ffn.experts_act","order":510,"layer":21,"block":"l21.moe_ffn","block_type":"moe_ffn","kind":"linear_attention","op":"silu_mul","roles":[],"inserted":false,"precision":{"weight":null,"activation":"bf16","output":"f32"},"steps":["compute f32"],"group":310,"kernel":"moe_shared_expert_fused_fp8::moe_expert_silu_down_shared_fp8","family":"moe_fp8_1row","point":null,"compute":"cuda_core","numerics":"reference","where":"device","state":null},{"id":"l21.moe_ffn.experts_down","order":511,"layer":21,"block":"l21.moe_ffn","block_type":"moe_ffn","kind":"linear_attention","op":"expert_down","roles":[],"inserted":false,"precision":{"weight":"fp8/block128x128","activation":"f32","output":"bf16"},"steps":["act f32","weight fp8/block128x128->f32","mma f32*f32","accumulate f32","scale f32"],"group":310,"kernel":"moe_shared_expert_fused_fp8::moe_expert_silu_down_shared_fp8","family":"moe_fp8_1row","point":null,"compute":"cuda_core","numerics":"reference","where":"device","state":null},{"id":"l21.moe_ffn.shared_act","order":512,"layer":21,"block":"l21.moe_ffn","block_type":"moe_ffn","kind":"linear_attention","op":"silu_mul","roles":[],"inserted":false,"precision":{"weight":null,"activation":"bf16","output":"f32"},"steps":["compute f32"],"group":310,"kernel":"moe_shared_expert_fused_fp8::moe_expert_silu_down_shared_fp8","family":"moe_fp8_1row","point":null,"compute":"cuda_core","numerics":"reference","where":"device","state":null},{"id":"l21.moe_ffn.shared_down","order":513,"layer":21,"block":"l21.moe_ffn","block_type":"moe_ffn","kind":"linear_attention","op":"linear:shared_down","roles":["shared_down"],"inserted":false,"precision":{"weight":"fp8/block128x128","activation":"f32","output":"bf16"},"steps":["act f32","weight fp8/block128x128->f32","mma f32*f32","accumulate f32","scale f32"],"group":310,"kernel":"moe_shared_expert_fused_fp8::moe_expert_silu_down_shared_fp8","family":"moe_fp8_1row","point":null,"compute":"cuda_core","numerics":"reference","where":"device","state":null},{"id":"l21.moe_ffn.shared_gate","order":514,"layer":21,"block":"l21.moe_ffn","block_type":"moe_ffn","kind":"linear_attention","op":"linear:shared_gate","roles":["shared_gate"],"inserted":false,"precision":{"weight":"bf16","activation":"bf16","output":"f32"},"steps":["act bf16","weight bf16->bf16","mma bf16*bf16","accumulate f32","scale none"],"group":311,"kernel":"moe_expert_gemv::moe_weighted_sum_blend","family":"moe_blend","point":{"shared_gate":"sigmoid"},"compute":"cuda_core","numerics":"reference","where":"device","state":null},{"id":"l21.moe_ffn.blend","order":515,"layer":21,"block":"l21.moe_ffn","block_type":"moe_ffn","kind":"linear_attention","op":"blend","roles":[],"inserted":false,"precision":{"weight":null,"activation":"bf16","output":"bf16"},"steps":["scatter f32","combine f32"],"group":311,"kernel":"moe_expert_gemv::moe_weighted_sum_blend","family":"moe_blend","point":{"shared_gate":"sigmoid"},"compute":"cuda_core","numerics":"reference","where":"device","state":null},{"id":"l21.moe_ffn.add","order":516,"layer":21,"block":"l21.moe_ffn","block_type":"moe_ffn","kind":"linear_attention","op":"residual_add","roles":[],"inserted":false,"precision":{"weight":null,"activation":"bf16","output":"bf16"},"steps":["compute f32"],"group":312,"kernel":"residual_add_rms_norm_exact::residual_add_rms_norm_exact","family":"rms_norm","point":null,"compute":"memory","numerics":"bit_identical(residual_add_rms_norm_exact_microtest)","where":"device","state":null},{"id":"l22.gdn.input_norm","order":517,"layer":22,"block":"l22.gdn","block_type":"gdn","kind":"linear_attention","op":"rms_norm","roles":[],"inserted":false,"precision":{"weight":null,"activation":"bf16","output":"bf16"},"steps":["compute f32"],"group":312,"kernel":"residual_add_rms_norm_exact::residual_add_rms_norm_exact","family":"rms_norm","point":null,"compute":"memory","numerics":"bit_identical(residual_add_rms_norm_exact_microtest)","where":"device","state":null},{"id":"l22.gdn.qkvz","order":518,"layer":22,"block":"l22.gdn","block_type":"gdn","kind":"linear_attention","op":"linear:qkvz","roles":["qkvz"],"inserted":false,"precision":{"weight":"fp8/block128x128","activation":"bf16","output":"bf16"},"steps":["act bf16","weight fp8/block128x128->bf16","mma bf16*bf16","accumulate f32","scale f32"],"group":313,"kernel":"w8a16_gemm_pipelined_m32::w8a16_gemm_pipelined_m32","family":"w8a16_gemm","point":null,"compute":"tensor_core:mma.sync.m16n8k16.bf16","numerics":"reference","where":"device","state":null},{"id":"l22.gdn.ba","order":519,"layer":22,"block":"l22.gdn","block_type":"gdn","kind":"linear_attention","op":"linear:ba","roles":["ba"],"inserted":false,"precision":{"weight":"bf16","activation":"bf16","output":"f32"},"steps":["act bf16","weight bf16->bf16","mma bf16*bf16","accumulate f32","scale none"],"group":314,"kernel":"ssm_preprocess::dense_gemv_ba_gates","family":"gdn_ba_gates_gemv","point":{"gate":"softplus","decay":"head"},"compute":"cuda_core","numerics":"reference","where":"device","state":null},{"id":"l22.gdn.gates","order":520,"layer":22,"block":"l22.gdn","block_type":"gdn","kind":"linear_attention","op":"gdn_gates","roles":[],"inserted":false,"precision":{"weight":null,"activation":"f32","output":"f32,f32"},"steps":["compute f32"],"group":314,"kernel":"ssm_preprocess::dense_gemv_ba_gates","family":"gdn_ba_gates_gemv","point":{"gate":"softplus","decay":"head"},"compute":"cuda_core","numerics":"reference","where":"device","state":null},{"id":"l22.gdn.conv","order":521,"layer":22,"block":"l22.gdn","block_type":"gdn","kind":"linear_attention","op":"conv1d_update","roles":[],"inserted":false,"precision":{"weight":null,"activation":"bf16","output":"f32"},"steps":["state f32","compute f32"],"group":315,"kernel":"causal_conv1d::causal_conv1d_update_l2norm_f32","family":"causal_conv1d_l2norm","point":null,"compute":"cuda_core","numerics":"reference","where":"device","state":"conv:update"},{"id":"l22.gdn.conv_ckpt","order":522,"layer":22,"block":"l22.gdn","block_type":"gdn","kind":"linear_attention","op":"state_snapshot","roles":[],"inserted":false,"precision":{"weight":null,"activation":"f32","output":null},"steps":["state f32"],"group":315,"kernel":"causal_conv1d::causal_conv1d_update_l2norm_f32","family":"causal_conv1d_l2norm","point":null,"compute":"cuda_core","numerics":"reference","where":"device","state":"conv:snapshot"},{"id":"l22.gdn.l2","order":523,"layer":22,"block":"l22.gdn","block_type":"gdn","kind":"linear_attention","op":"l2_norm","roles":[],"inserted":false,"precision":{"weight":null,"activation":"f32","output":"f32"},"steps":["compute f32"],"group":315,"kernel":"causal_conv1d::causal_conv1d_update_l2norm_f32","family":"causal_conv1d_l2norm","point":null,"compute":"cuda_core","numerics":"reference","where":"device","state":null},{"id":"l22.gdn.recur","order":524,"layer":22,"block":"l22.gdn","block_type":"gdn","kind":"linear_attention","op":"gdn_recurrence","roles":[],"inserted":false,"precision":{"weight":null,"activation":"f32","output":"f32"},"steps":["state f32","compute f32"],"group":316,"kernel":"gated_delta_rule::gated_delta_rule_decode_f32","family":"gdn_recurrence","point":{"decay":"head","gate_act":"silu"},"compute":"cuda_core","numerics":"reference","where":"device","state":"h:update"},{"id":"l22.gdn.out_norm","order":525,"layer":22,"block":"l22.gdn","block_type":"gdn","kind":"linear_attention","op":"gated_rms_norm","roles":[],"inserted":false,"precision":{"weight":null,"activation":"f32","output":"bf16"},"steps":["compute f32"],"group":317,"kernel":"norm::gated_rms_norm_f32_input","family":"gated_rms_norm","point":null,"compute":"memory","numerics":"reference","where":"device","state":null},{"id":"l22.gdn.out","order":526,"layer":22,"block":"l22.gdn","block_type":"gdn","kind":"linear_attention","op":"linear:gdn_out","roles":["gdn_out"],"inserted":false,"precision":{"weight":"fp8/block128x128","activation":"bf16","output":"bf16"},"steps":["act bf16","weight fp8/block128x128->bf16","mma bf16*bf16","accumulate f32","scale f32"],"group":318,"kernel":"w8a16_gemm_pipelined_m32::w8a16_gemm_pipelined_m32","family":"w8a16_gemm","point":null,"compute":"tensor_core:mma.sync.m16n8k16.bf16","numerics":"reference","where":"device","state":null},{"id":"l22.gdn.add","order":527,"layer":22,"block":"l22.gdn","block_type":"gdn","kind":"linear_attention","op":"residual_add","roles":[],"inserted":false,"precision":{"weight":null,"activation":"bf16","output":"bf16"},"steps":["compute f32"],"group":319,"kernel":"norm::residual_add_rms_norm","family":"rms_norm","point":null,"compute":"memory","numerics":"reference","where":"device","state":null},{"id":"l22.moe_ffn.post_norm","order":528,"layer":22,"block":"l22.moe_ffn","block_type":"moe_ffn","kind":"linear_attention","op":"rms_norm","roles":[],"inserted":false,"precision":{"weight":null,"activation":"f32","output":"bf16"},"steps":["compute f32"],"group":319,"kernel":"norm::residual_add_rms_norm","family":"rms_norm","point":null,"compute":"memory","numerics":"reference","where":"device","state":null},{"id":"l22.moe_ffn.router","order":529,"layer":22,"block":"l22.moe_ffn","block_type":"moe_ffn","kind":"linear_attention","op":"router","roles":[],"inserted":false,"precision":{"weight":"bf16","activation":"bf16","output":"bf16"},"steps":["act bf16","weight bf16->bf16","mma bf16*bf16","accumulate f32","scale none"],"group":320,"kernel":"gemv::dense_gemv_bf16","family":"dense_bf16","point":null,"compute":"cuda_core","numerics":"reference","where":"device","state":null},{"id":"l22.moe_ffn.top_k","order":530,"layer":22,"block":"l22.moe_ffn","block_type":"moe_ffn","kind":"linear_attention","op":"top_k","roles":[],"inserted":false,"precision":{"weight":null,"activation":"bf16","output":"f32,i32"},"steps":["score f32"],"group":321,"kernel":"moe_topk::moe_topk_softmax","family":"moe_topk","point":null,"compute":"memory","numerics":"reference","where":"device","state":null},{"id":"l22.moe_ffn.experts_gate_up","order":531,"layer":22,"block":"l22.moe_ffn","block_type":"moe_ffn","kind":"linear_attention","op":"expert_gate_up","roles":[],"inserted":false,"precision":{"weight":"fp8/block128x128","activation":"bf16","output":"bf16"},"steps":["gather bf16","act bf16","weight fp8/block128x128->bf16","mma bf16*bf16","accumulate f32","scale f32"],"group":322,"kernel":"moe_shared_expert_fused_fp8::moe_expert_gate_up_shared_fp8","family":"moe_fp8_1row","point":null,"compute":"cuda_core","numerics":"reference","where":"device","state":null},{"id":"l22.moe_ffn.shared_gate_up","order":532,"layer":22,"block":"l22.moe_ffn","block_type":"moe_ffn","kind":"linear_attention","op":"linear:shared_gate_up","roles":["shared_gate_up"],"inserted":false,"precision":{"weight":"fp8/block128x128","activation":"bf16","output":"bf16"},"steps":["act bf16","weight fp8/block128x128->bf16","mma bf16*bf16","accumulate f32","scale f32"],"group":322,"kernel":"moe_shared_expert_fused_fp8::moe_expert_gate_up_shared_fp8","family":"moe_fp8_1row","point":null,"compute":"cuda_core","numerics":"reference","where":"device","state":null},{"id":"l22.moe_ffn.experts_act","order":533,"layer":22,"block":"l22.moe_ffn","block_type":"moe_ffn","kind":"linear_attention","op":"silu_mul","roles":[],"inserted":false,"precision":{"weight":null,"activation":"bf16","output":"f32"},"steps":["compute f32"],"group":323,"kernel":"moe_shared_expert_fused_fp8::moe_expert_silu_down_shared_fp8","family":"moe_fp8_1row","point":null,"compute":"cuda_core","numerics":"reference","where":"device","state":null},{"id":"l22.moe_ffn.experts_down","order":534,"layer":22,"block":"l22.moe_ffn","block_type":"moe_ffn","kind":"linear_attention","op":"expert_down","roles":[],"inserted":false,"precision":{"weight":"fp8/block128x128","activation":"f32","output":"bf16"},"steps":["act f32","weight fp8/block128x128->f32","mma f32*f32","accumulate f32","scale f32"],"group":323,"kernel":"moe_shared_expert_fused_fp8::moe_expert_silu_down_shared_fp8","family":"moe_fp8_1row","point":null,"compute":"cuda_core","numerics":"reference","where":"device","state":null},{"id":"l22.moe_ffn.shared_act","order":535,"layer":22,"block":"l22.moe_ffn","block_type":"moe_ffn","kind":"linear_attention","op":"silu_mul","roles":[],"inserted":false,"precision":{"weight":null,"activation":"bf16","output":"f32"},"steps":["compute f32"],"group":323,"kernel":"moe_shared_expert_fused_fp8::moe_expert_silu_down_shared_fp8","family":"moe_fp8_1row","point":null,"compute":"cuda_core","numerics":"reference","where":"device","state":null},{"id":"l22.moe_ffn.shared_down","order":536,"layer":22,"block":"l22.moe_ffn","block_type":"moe_ffn","kind":"linear_attention","op":"linear:shared_down","roles":["shared_down"],"inserted":false,"precision":{"weight":"fp8/block128x128","activation":"f32","output":"bf16"},"steps":["act f32","weight fp8/block128x128->f32","mma f32*f32","accumulate f32","scale f32"],"group":323,"kernel":"moe_shared_expert_fused_fp8::moe_expert_silu_down_shared_fp8","family":"moe_fp8_1row","point":null,"compute":"cuda_core","numerics":"reference","where":"device","state":null},{"id":"l22.moe_ffn.shared_gate","order":537,"layer":22,"block":"l22.moe_ffn","block_type":"moe_ffn","kind":"linear_attention","op":"linear:shared_gate","roles":["shared_gate"],"inserted":false,"precision":{"weight":"bf16","activation":"bf16","output":"f32"},"steps":["act bf16","weight bf16->bf16","mma bf16*bf16","accumulate f32","scale none"],"group":324,"kernel":"moe_expert_gemv::moe_weighted_sum_blend","family":"moe_blend","point":{"shared_gate":"sigmoid"},"compute":"cuda_core","numerics":"reference","where":"device","state":null},{"id":"l22.moe_ffn.blend","order":538,"layer":22,"block":"l22.moe_ffn","block_type":"moe_ffn","kind":"linear_attention","op":"blend","roles":[],"inserted":false,"precision":{"weight":null,"activation":"bf16","output":"bf16"},"steps":["scatter f32","combine f32"],"group":324,"kernel":"moe_expert_gemv::moe_weighted_sum_blend","family":"moe_blend","point":{"shared_gate":"sigmoid"},"compute":"cuda_core","numerics":"reference","where":"device","state":null},{"id":"l22.moe_ffn.add","order":539,"layer":22,"block":"l22.moe_ffn","block_type":"moe_ffn","kind":"linear_attention","op":"residual_add","roles":[],"inserted":false,"precision":{"weight":null,"activation":"bf16","output":"bf16"},"steps":["compute f32"],"group":325,"kernel":"residual_add_rms_norm_exact::residual_add_rms_norm_exact","family":"rms_norm","point":null,"compute":"memory","numerics":"bit_identical(residual_add_rms_norm_exact_microtest)","where":"device","state":null},{"id":"l23.attn.input_norm","order":540,"layer":23,"block":"l23.attn","block_type":"attn","kind":"full_attention","op":"rms_norm","roles":[],"inserted":false,"precision":{"weight":null,"activation":"bf16","output":"bf16"},"steps":["compute f32"],"group":325,"kernel":"residual_add_rms_norm_exact::residual_add_rms_norm_exact","family":"rms_norm","point":null,"compute":"memory","numerics":"bit_identical(residual_add_rms_norm_exact_microtest)","where":"device","state":null},{"id":"l23.attn.q","order":541,"layer":23,"block":"l23.attn","block_type":"attn","kind":"full_attention","op":"linear:q","roles":["q"],"inserted":false,"precision":{"weight":"fp8/block128x128","activation":"bf16","output":"bf16"},"steps":["act bf16","weight fp8/block128x128->bf16","mma bf16*bf16","accumulate f32","scale f32"],"group":326,"kernel":"w8a16_gemm_pipelined_m32::w8a16_gemm_pipelined_m32","family":"w8a16_gemm","point":null,"compute":"tensor_core:mma.sync.m16n8k16.bf16","numerics":"reference","where":"device","state":null},{"id":"l23.attn.q_split","order":542,"layer":23,"block":"l23.attn","block_type":"attn","kind":"full_attention","op":"split","roles":[],"inserted":false,"precision":{"weight":null,"activation":"bf16","output":"bf16,bf16"},"steps":["move bf16"],"group":327,"kernel":"ssm_preprocess::deinterleave_qg","family":"deinterleave_qg","point":{"layout":"interleaved"},"compute":"memory","numerics":"reference","where":"device","state":null},{"id":"l23.attn.k","order":543,"layer":23,"block":"l23.attn","block_type":"attn","kind":"full_attention","op":"linear:k","roles":["k"],"inserted":false,"precision":{"weight":"fp8/block128x128","activation":"bf16","output":"bf16"},"steps":["act bf16","weight fp8/block128x128->bf16","mma bf16*bf16","accumulate f32","scale f32"],"group":328,"kernel":"w8a16_gemm_pipelined_m32::w8a16_gemm_pipelined_m32","family":"w8a16_gemm","point":null,"compute":"tensor_core:mma.sync.m16n8k16.bf16","numerics":"reference","where":"device","state":null},{"id":"l23.attn.v","order":544,"layer":23,"block":"l23.attn","block_type":"attn","kind":"full_attention","op":"linear:v","roles":["v"],"inserted":false,"precision":{"weight":"fp8/block128x128","activation":"bf16","output":"bf16"},"steps":["act bf16","weight fp8/block128x128->bf16","mma bf16*bf16","accumulate f32","scale f32"],"group":329,"kernel":"w8a16_gemm_pipelined_m32::w8a16_gemm_pipelined_m32","family":"w8a16_gemm","point":null,"compute":"tensor_core:mma.sync.m16n8k16.bf16","numerics":"reference","where":"device","state":null},{"id":"l23.attn.q_norm","order":545,"layer":23,"block":"l23.attn","block_type":"attn","kind":"full_attention","op":"qk_norm","roles":[],"inserted":false,"precision":{"weight":null,"activation":"bf16","output":"bf16"},"steps":["compute f32"],"group":330,"kernel":"norm::rms_norm","family":"rms_norm","point":null,"compute":"memory","numerics":"reference","where":"device","state":null},{"id":"l23.attn.k_norm","order":546,"layer":23,"block":"l23.attn","block_type":"attn","kind":"full_attention","op":"qk_norm","roles":[],"inserted":false,"precision":{"weight":null,"activation":"bf16","output":"bf16"},"steps":["compute f32"],"group":331,"kernel":"norm::rms_norm","family":"rms_norm","point":null,"compute":"memory","numerics":"reference","where":"device","state":null},{"id":"l23.attn.rope","order":547,"layer":23,"block":"l23.attn","block_type":"attn","kind":"full_attention","op":"rope","roles":[],"inserted":false,"precision":{"weight":null,"activation":"bf16","output":"bf16,bf16"},"steps":["compute f32"],"group":332,"kernel":"rope_mrope_interleaved::rope_forward_mrope_interleaved","family":"rope","point":null,"compute":"memory","numerics":"reference","where":"device","state":null},{"id":"l23.attn.kv_write","order":548,"layer":23,"block":"l23.attn","block_type":"attn","kind":"full_attention","op":"kv_write","roles":[],"inserted":false,"precision":{"weight":null,"activation":"bf16","output":null},"steps":["cache bf16"],"group":333,"kernel":"reshape_and_cache::reshape_and_cache_flash","family":"kv_write","point":null,"compute":"memory","numerics":"reference","where":"device","state":"k:write,v:write"},{"id":"l23.attn.attend","order":549,"layer":23,"block":"l23.attn","block_type":"attn","kind":"full_attention","op":"paged_attention","roles":[],"inserted":false,"precision":{"weight":null,"activation":"bf16","output":"bf16"},"steps":["cache bf16","scores f32","softmax f32","accumulate f32"],"group":334,"kernel":"paged_decode::paged_decode_attn","family":"paged_decode_attn","point":null,"compute":"cuda_core","numerics":"reference","where":"device","state":"k:read,v:read"},{"id":"l23.attn.gate_mul","order":550,"layer":23,"block":"l23.attn","block_type":"attn","kind":"full_attention","op":"sigmoid_gate_mul","roles":[],"inserted":false,"precision":{"weight":null,"activation":"bf16","output":"bf16"},"steps":["compute f32"],"group":335,"kernel":"residual_add::sigmoid_gate_mul","family":"sigmoid_gate_mul","point":null,"compute":"memory","numerics":"reference","where":"device","state":null},{"id":"l23.attn.o","order":551,"layer":23,"block":"l23.attn","block_type":"attn","kind":"full_attention","op":"linear:o","roles":["o"],"inserted":false,"precision":{"weight":"fp8/block128x128","activation":"bf16","output":"bf16"},"steps":["act bf16","weight fp8/block128x128->bf16","mma bf16*bf16","accumulate f32","scale f32"],"group":336,"kernel":"w8a16_gemm_pipelined_m32::w8a16_gemm_pipelined_m32","family":"w8a16_gemm","point":null,"compute":"tensor_core:mma.sync.m16n8k16.bf16","numerics":"reference","where":"device","state":null},{"id":"l23.attn.add","order":552,"layer":23,"block":"l23.attn","block_type":"attn","kind":"full_attention","op":"residual_add","roles":[],"inserted":false,"precision":{"weight":null,"activation":"bf16","output":"bf16"},"steps":["compute f32"],"group":337,"kernel":"norm::residual_add_rms_norm","family":"rms_norm","point":null,"compute":"memory","numerics":"reference","where":"device","state":null},{"id":"l23.moe_ffn.post_norm","order":553,"layer":23,"block":"l23.moe_ffn","block_type":"moe_ffn","kind":"full_attention","op":"rms_norm","roles":[],"inserted":false,"precision":{"weight":null,"activation":"f32","output":"bf16"},"steps":["compute f32"],"group":337,"kernel":"norm::residual_add_rms_norm","family":"rms_norm","point":null,"compute":"memory","numerics":"reference","where":"device","state":null},{"id":"l23.moe_ffn.router","order":554,"layer":23,"block":"l23.moe_ffn","block_type":"moe_ffn","kind":"full_attention","op":"router","roles":[],"inserted":false,"precision":{"weight":"bf16","activation":"bf16","output":"bf16"},"steps":["act bf16","weight bf16->bf16","mma bf16*bf16","accumulate f32","scale none"],"group":338,"kernel":"gemv::dense_gemv_bf16","family":"dense_bf16","point":null,"compute":"cuda_core","numerics":"reference","where":"device","state":null},{"id":"l23.moe_ffn.top_k","order":555,"layer":23,"block":"l23.moe_ffn","block_type":"moe_ffn","kind":"full_attention","op":"top_k","roles":[],"inserted":false,"precision":{"weight":null,"activation":"bf16","output":"f32,i32"},"steps":["score f32"],"group":339,"kernel":"moe_topk::moe_topk_softmax","family":"moe_topk","point":null,"compute":"memory","numerics":"reference","where":"device","state":null},{"id":"l23.moe_ffn.experts_gate_up","order":556,"layer":23,"block":"l23.moe_ffn","block_type":"moe_ffn","kind":"full_attention","op":"expert_gate_up","roles":[],"inserted":false,"precision":{"weight":"fp8/block128x128","activation":"bf16","output":"bf16"},"steps":["gather bf16","act bf16","weight fp8/block128x128->bf16","mma bf16*bf16","accumulate f32","scale f32"],"group":340,"kernel":"moe_shared_expert_fused_fp8::moe_expert_gate_up_shared_fp8","family":"moe_fp8_1row","point":null,"compute":"cuda_core","numerics":"reference","where":"device","state":null},{"id":"l23.moe_ffn.shared_gate_up","order":557,"layer":23,"block":"l23.moe_ffn","block_type":"moe_ffn","kind":"full_attention","op":"linear:shared_gate_up","roles":["shared_gate_up"],"inserted":false,"precision":{"weight":"fp8/block128x128","activation":"bf16","output":"bf16"},"steps":["act bf16","weight fp8/block128x128->bf16","mma bf16*bf16","accumulate f32","scale f32"],"group":340,"kernel":"moe_shared_expert_fused_fp8::moe_expert_gate_up_shared_fp8","family":"moe_fp8_1row","point":null,"compute":"cuda_core","numerics":"reference","where":"device","state":null},{"id":"l23.moe_ffn.experts_act","order":558,"layer":23,"block":"l23.moe_ffn","block_type":"moe_ffn","kind":"full_attention","op":"silu_mul","roles":[],"inserted":false,"precision":{"weight":null,"activation":"bf16","output":"f32"},"steps":["compute f32"],"group":341,"kernel":"moe_shared_expert_fused_fp8::moe_expert_silu_down_shared_fp8","family":"moe_fp8_1row","point":null,"compute":"cuda_core","numerics":"reference","where":"device","state":null},{"id":"l23.moe_ffn.experts_down","order":559,"layer":23,"block":"l23.moe_ffn","block_type":"moe_ffn","kind":"full_attention","op":"expert_down","roles":[],"inserted":false,"precision":{"weight":"fp8/block128x128","activation":"f32","output":"bf16"},"steps":["act f32","weight fp8/block128x128->f32","mma f32*f32","accumulate f32","scale f32"],"group":341,"kernel":"moe_shared_expert_fused_fp8::moe_expert_silu_down_shared_fp8","family":"moe_fp8_1row","point":null,"compute":"cuda_core","numerics":"reference","where":"device","state":null},{"id":"l23.moe_ffn.shared_act","order":560,"layer":23,"block":"l23.moe_ffn","block_type":"moe_ffn","kind":"full_attention","op":"silu_mul","roles":[],"inserted":false,"precision":{"weight":null,"activation":"bf16","output":"f32"},"steps":["compute f32"],"group":341,"kernel":"moe_shared_expert_fused_fp8::moe_expert_silu_down_shared_fp8","family":"moe_fp8_1row","point":null,"compute":"cuda_core","numerics":"reference","where":"device","state":null},{"id":"l23.moe_ffn.shared_down","order":561,"layer":23,"block":"l23.moe_ffn","block_type":"moe_ffn","kind":"full_attention","op":"linear:shared_down","roles":["shared_down"],"inserted":false,"precision":{"weight":"fp8/block128x128","activation":"f32","output":"bf16"},"steps":["act f32","weight fp8/block128x128->f32","mma f32*f32","accumulate f32","scale f32"],"group":341,"kernel":"moe_shared_expert_fused_fp8::moe_expert_silu_down_shared_fp8","family":"moe_fp8_1row","point":null,"compute":"cuda_core","numerics":"reference","where":"device","state":null},{"id":"l23.moe_ffn.shared_gate","order":562,"layer":23,"block":"l23.moe_ffn","block_type":"moe_ffn","kind":"full_attention","op":"linear:shared_gate","roles":["shared_gate"],"inserted":false,"precision":{"weight":"bf16","activation":"bf16","output":"f32"},"steps":["act bf16","weight bf16->bf16","mma bf16*bf16","accumulate f32","scale none"],"group":342,"kernel":"moe_expert_gemv::moe_weighted_sum_blend","family":"moe_blend","point":{"shared_gate":"sigmoid"},"compute":"cuda_core","numerics":"reference","where":"device","state":null},{"id":"l23.moe_ffn.blend","order":563,"layer":23,"block":"l23.moe_ffn","block_type":"moe_ffn","kind":"full_attention","op":"blend","roles":[],"inserted":false,"precision":{"weight":null,"activation":"bf16","output":"bf16"},"steps":["scatter f32","combine f32"],"group":342,"kernel":"moe_expert_gemv::moe_weighted_sum_blend","family":"moe_blend","point":{"shared_gate":"sigmoid"},"compute":"cuda_core","numerics":"reference","where":"device","state":null},{"id":"l23.moe_ffn.add","order":564,"layer":23,"block":"l23.moe_ffn","block_type":"moe_ffn","kind":"full_attention","op":"residual_add","roles":[],"inserted":false,"precision":{"weight":null,"activation":"bf16","output":"bf16"},"steps":["compute f32"],"group":343,"kernel":"residual_add_rms_norm_exact::residual_add_rms_norm_exact","family":"rms_norm","point":null,"compute":"memory","numerics":"bit_identical(residual_add_rms_norm_exact_microtest)","where":"device","state":null},{"id":"l24.gdn.input_norm","order":565,"layer":24,"block":"l24.gdn","block_type":"gdn","kind":"linear_attention","op":"rms_norm","roles":[],"inserted":false,"precision":{"weight":null,"activation":"bf16","output":"bf16"},"steps":["compute f32"],"group":343,"kernel":"residual_add_rms_norm_exact::residual_add_rms_norm_exact","family":"rms_norm","point":null,"compute":"memory","numerics":"bit_identical(residual_add_rms_norm_exact_microtest)","where":"device","state":null},{"id":"l24.gdn.qkvz","order":566,"layer":24,"block":"l24.gdn","block_type":"gdn","kind":"linear_attention","op":"linear:qkvz","roles":["qkvz"],"inserted":false,"precision":{"weight":"fp8/block128x128","activation":"bf16","output":"bf16"},"steps":["act bf16","weight fp8/block128x128->bf16","mma bf16*bf16","accumulate f32","scale f32"],"group":344,"kernel":"w8a16_gemm_pipelined_m32::w8a16_gemm_pipelined_m32","family":"w8a16_gemm","point":null,"compute":"tensor_core:mma.sync.m16n8k16.bf16","numerics":"reference","where":"device","state":null},{"id":"l24.gdn.ba","order":567,"layer":24,"block":"l24.gdn","block_type":"gdn","kind":"linear_attention","op":"linear:ba","roles":["ba"],"inserted":false,"precision":{"weight":"bf16","activation":"bf16","output":"f32"},"steps":["act bf16","weight bf16->bf16","mma bf16*bf16","accumulate f32","scale none"],"group":345,"kernel":"ssm_preprocess::dense_gemv_ba_gates","family":"gdn_ba_gates_gemv","point":{"gate":"softplus","decay":"head"},"compute":"cuda_core","numerics":"reference","where":"device","state":null},{"id":"l24.gdn.gates","order":568,"layer":24,"block":"l24.gdn","block_type":"gdn","kind":"linear_attention","op":"gdn_gates","roles":[],"inserted":false,"precision":{"weight":null,"activation":"f32","output":"f32,f32"},"steps":["compute f32"],"group":345,"kernel":"ssm_preprocess::dense_gemv_ba_gates","family":"gdn_ba_gates_gemv","point":{"gate":"softplus","decay":"head"},"compute":"cuda_core","numerics":"reference","where":"device","state":null},{"id":"l24.gdn.conv","order":569,"layer":24,"block":"l24.gdn","block_type":"gdn","kind":"linear_attention","op":"conv1d_update","roles":[],"inserted":false,"precision":{"weight":null,"activation":"bf16","output":"f32"},"steps":["state f32","compute f32"],"group":346,"kernel":"causal_conv1d::causal_conv1d_update_l2norm_f32","family":"causal_conv1d_l2norm","point":null,"compute":"cuda_core","numerics":"reference","where":"device","state":"conv:update"},{"id":"l24.gdn.conv_ckpt","order":570,"layer":24,"block":"l24.gdn","block_type":"gdn","kind":"linear_attention","op":"state_snapshot","roles":[],"inserted":false,"precision":{"weight":null,"activation":"f32","output":null},"steps":["state f32"],"group":346,"kernel":"causal_conv1d::causal_conv1d_update_l2norm_f32","family":"causal_conv1d_l2norm","point":null,"compute":"cuda_core","numerics":"reference","where":"device","state":"conv:snapshot"},{"id":"l24.gdn.l2","order":571,"layer":24,"block":"l24.gdn","block_type":"gdn","kind":"linear_attention","op":"l2_norm","roles":[],"inserted":false,"precision":{"weight":null,"activation":"f32","output":"f32"},"steps":["compute f32"],"group":346,"kernel":"causal_conv1d::causal_conv1d_update_l2norm_f32","family":"causal_conv1d_l2norm","point":null,"compute":"cuda_core","numerics":"reference","where":"device","state":null},{"id":"l24.gdn.recur","order":572,"layer":24,"block":"l24.gdn","block_type":"gdn","kind":"linear_attention","op":"gdn_recurrence","roles":[],"inserted":false,"precision":{"weight":null,"activation":"f32","output":"f32"},"steps":["state f32","compute f32"],"group":347,"kernel":"gated_delta_rule::gated_delta_rule_decode_f32","family":"gdn_recurrence","point":{"decay":"head","gate_act":"silu"},"compute":"cuda_core","numerics":"reference","where":"device","state":"h:update"},{"id":"l24.gdn.out_norm","order":573,"layer":24,"block":"l24.gdn","block_type":"gdn","kind":"linear_attention","op":"gated_rms_norm","roles":[],"inserted":false,"precision":{"weight":null,"activation":"f32","output":"bf16"},"steps":["compute f32"],"group":348,"kernel":"norm::gated_rms_norm_f32_input","family":"gated_rms_norm","point":null,"compute":"memory","numerics":"reference","where":"device","state":null},{"id":"l24.gdn.out","order":574,"layer":24,"block":"l24.gdn","block_type":"gdn","kind":"linear_attention","op":"linear:gdn_out","roles":["gdn_out"],"inserted":false,"precision":{"weight":"fp8/block128x128","activation":"bf16","output":"bf16"},"steps":["act bf16","weight fp8/block128x128->bf16","mma bf16*bf16","accumulate f32","scale f32"],"group":349,"kernel":"w8a16_gemm_pipelined_m32::w8a16_gemm_pipelined_m32","family":"w8a16_gemm","point":null,"compute":"tensor_core:mma.sync.m16n8k16.bf16","numerics":"reference","where":"device","state":null},{"id":"l24.gdn.add","order":575,"layer":24,"block":"l24.gdn","block_type":"gdn","kind":"linear_attention","op":"residual_add","roles":[],"inserted":false,"precision":{"weight":null,"activation":"bf16","output":"bf16"},"steps":["compute f32"],"group":350,"kernel":"norm::residual_add_rms_norm","family":"rms_norm","point":null,"compute":"memory","numerics":"reference","where":"device","state":null},{"id":"l24.moe_ffn.post_norm","order":576,"layer":24,"block":"l24.moe_ffn","block_type":"moe_ffn","kind":"linear_attention","op":"rms_norm","roles":[],"inserted":false,"precision":{"weight":null,"activation":"f32","output":"bf16"},"steps":["compute f32"],"group":350,"kernel":"norm::residual_add_rms_norm","family":"rms_norm","point":null,"compute":"memory","numerics":"reference","where":"device","state":null},{"id":"l24.moe_ffn.router","order":577,"layer":24,"block":"l24.moe_ffn","block_type":"moe_ffn","kind":"linear_attention","op":"router","roles":[],"inserted":false,"precision":{"weight":"bf16","activation":"bf16","output":"bf16"},"steps":["act bf16","weight bf16->bf16","mma bf16*bf16","accumulate f32","scale none"],"group":351,"kernel":"gemv::dense_gemv_bf16","family":"dense_bf16","point":null,"compute":"cuda_core","numerics":"reference","where":"device","state":null},{"id":"l24.moe_ffn.top_k","order":578,"layer":24,"block":"l24.moe_ffn","block_type":"moe_ffn","kind":"linear_attention","op":"top_k","roles":[],"inserted":false,"precision":{"weight":null,"activation":"bf16","output":"f32,i32"},"steps":["score f32"],"group":352,"kernel":"moe_topk::moe_topk_softmax","family":"moe_topk","point":null,"compute":"memory","numerics":"reference","where":"device","state":null},{"id":"l24.moe_ffn.experts_gate_up","order":579,"layer":24,"block":"l24.moe_ffn","block_type":"moe_ffn","kind":"linear_attention","op":"expert_gate_up","roles":[],"inserted":false,"precision":{"weight":"fp8/block128x128","activation":"bf16","output":"bf16"},"steps":["gather bf16","act bf16","weight fp8/block128x128->bf16","mma bf16*bf16","accumulate f32","scale f32"],"group":353,"kernel":"moe_shared_expert_fused_fp8::moe_expert_gate_up_shared_fp8","family":"moe_fp8_1row","point":null,"compute":"cuda_core","numerics":"reference","where":"device","state":null},{"id":"l24.moe_ffn.shared_gate_up","order":580,"layer":24,"block":"l24.moe_ffn","block_type":"moe_ffn","kind":"linear_attention","op":"linear:shared_gate_up","roles":["shared_gate_up"],"inserted":false,"precision":{"weight":"fp8/block128x128","activation":"bf16","output":"bf16"},"steps":["act bf16","weight fp8/block128x128->bf16","mma bf16*bf16","accumulate f32","scale f32"],"group":353,"kernel":"moe_shared_expert_fused_fp8::moe_expert_gate_up_shared_fp8","family":"moe_fp8_1row","point":null,"compute":"cuda_core","numerics":"reference","where":"device","state":null},{"id":"l24.moe_ffn.experts_act","order":581,"layer":24,"block":"l24.moe_ffn","block_type":"moe_ffn","kind":"linear_attention","op":"silu_mul","roles":[],"inserted":false,"precision":{"weight":null,"activation":"bf16","output":"f32"},"steps":["compute f32"],"group":354,"kernel":"moe_shared_expert_fused_fp8::moe_expert_silu_down_shared_fp8","family":"moe_fp8_1row","point":null,"compute":"cuda_core","numerics":"reference","where":"device","state":null},{"id":"l24.moe_ffn.experts_down","order":582,"layer":24,"block":"l24.moe_ffn","block_type":"moe_ffn","kind":"linear_attention","op":"expert_down","roles":[],"inserted":false,"precision":{"weight":"fp8/block128x128","activation":"f32","output":"bf16"},"steps":["act f32","weight fp8/block128x128->f32","mma f32*f32","accumulate f32","scale f32"],"group":354,"kernel":"moe_shared_expert_fused_fp8::moe_expert_silu_down_shared_fp8","family":"moe_fp8_1row","point":null,"compute":"cuda_core","numerics":"reference","where":"device","state":null},{"id":"l24.moe_ffn.shared_act","order":583,"layer":24,"block":"l24.moe_ffn","block_type":"moe_ffn","kind":"linear_attention","op":"silu_mul","roles":[],"inserted":false,"precision":{"weight":null,"activation":"bf16","output":"f32"},"steps":["compute f32"],"group":354,"kernel":"moe_shared_expert_fused_fp8::moe_expert_silu_down_shared_fp8","family":"moe_fp8_1row","point":null,"compute":"cuda_core","numerics":"reference","where":"device","state":null},{"id":"l24.moe_ffn.shared_down","order":584,"layer":24,"block":"l24.moe_ffn","block_type":"moe_ffn","kind":"linear_attention","op":"linear:shared_down","roles":["shared_down"],"inserted":false,"precision":{"weight":"fp8/block128x128","activation":"f32","output":"bf16"},"steps":["act f32","weight fp8/block128x128->f32","mma f32*f32","accumulate f32","scale f32"],"group":354,"kernel":"moe_shared_expert_fused_fp8::moe_expert_silu_down_shared_fp8","family":"moe_fp8_1row","point":null,"compute":"cuda_core","numerics":"reference","where":"device","state":null},{"id":"l24.moe_ffn.shared_gate","order":585,"layer":24,"block":"l24.moe_ffn","block_type":"moe_ffn","kind":"linear_attention","op":"linear:shared_gate","roles":["shared_gate"],"inserted":false,"precision":{"weight":"bf16","activation":"bf16","output":"f32"},"steps":["act bf16","weight bf16->bf16","mma bf16*bf16","accumulate f32","scale none"],"group":355,"kernel":"moe_expert_gemv::moe_weighted_sum_blend","family":"moe_blend","point":{"shared_gate":"sigmoid"},"compute":"cuda_core","numerics":"reference","where":"device","state":null},{"id":"l24.moe_ffn.blend","order":586,"layer":24,"block":"l24.moe_ffn","block_type":"moe_ffn","kind":"linear_attention","op":"blend","roles":[],"inserted":false,"precision":{"weight":null,"activation":"bf16","output":"bf16"},"steps":["scatter f32","combine f32"],"group":355,"kernel":"moe_expert_gemv::moe_weighted_sum_blend","family":"moe_blend","point":{"shared_gate":"sigmoid"},"compute":"cuda_core","numerics":"reference","where":"device","state":null},{"id":"l24.moe_ffn.add","order":587,"layer":24,"block":"l24.moe_ffn","block_type":"moe_ffn","kind":"linear_attention","op":"residual_add","roles":[],"inserted":false,"precision":{"weight":null,"activation":"bf16","output":"bf16"},"steps":["compute f32"],"group":356,"kernel":"residual_add_rms_norm_exact::residual_add_rms_norm_exact","family":"rms_norm","point":null,"compute":"memory","numerics":"bit_identical(residual_add_rms_norm_exact_microtest)","where":"device","state":null},{"id":"l25.gdn.input_norm","order":588,"layer":25,"block":"l25.gdn","block_type":"gdn","kind":"linear_attention","op":"rms_norm","roles":[],"inserted":false,"precision":{"weight":null,"activation":"bf16","output":"bf16"},"steps":["compute f32"],"group":356,"kernel":"residual_add_rms_norm_exact::residual_add_rms_norm_exact","family":"rms_norm","point":null,"compute":"memory","numerics":"bit_identical(residual_add_rms_norm_exact_microtest)","where":"device","state":null},{"id":"l25.gdn.qkvz","order":589,"layer":25,"block":"l25.gdn","block_type":"gdn","kind":"linear_attention","op":"linear:qkvz","roles":["qkvz"],"inserted":false,"precision":{"weight":"fp8/block128x128","activation":"bf16","output":"bf16"},"steps":["act bf16","weight fp8/block128x128->bf16","mma bf16*bf16","accumulate f32","scale f32"],"group":357,"kernel":"w8a16_gemm_pipelined_m32::w8a16_gemm_pipelined_m32","family":"w8a16_gemm","point":null,"compute":"tensor_core:mma.sync.m16n8k16.bf16","numerics":"reference","where":"device","state":null},{"id":"l25.gdn.ba","order":590,"layer":25,"block":"l25.gdn","block_type":"gdn","kind":"linear_attention","op":"linear:ba","roles":["ba"],"inserted":false,"precision":{"weight":"bf16","activation":"bf16","output":"f32"},"steps":["act bf16","weight bf16->bf16","mma bf16*bf16","accumulate f32","scale none"],"group":358,"kernel":"ssm_preprocess::dense_gemv_ba_gates","family":"gdn_ba_gates_gemv","point":{"gate":"softplus","decay":"head"},"compute":"cuda_core","numerics":"reference","where":"device","state":null},{"id":"l25.gdn.gates","order":591,"layer":25,"block":"l25.gdn","block_type":"gdn","kind":"linear_attention","op":"gdn_gates","roles":[],"inserted":false,"precision":{"weight":null,"activation":"f32","output":"f32,f32"},"steps":["compute f32"],"group":358,"kernel":"ssm_preprocess::dense_gemv_ba_gates","family":"gdn_ba_gates_gemv","point":{"gate":"softplus","decay":"head"},"compute":"cuda_core","numerics":"reference","where":"device","state":null},{"id":"l25.gdn.conv","order":592,"layer":25,"block":"l25.gdn","block_type":"gdn","kind":"linear_attention","op":"conv1d_update","roles":[],"inserted":false,"precision":{"weight":null,"activation":"bf16","output":"f32"},"steps":["state f32","compute f32"],"group":359,"kernel":"causal_conv1d::causal_conv1d_update_l2norm_f32","family":"causal_conv1d_l2norm","point":null,"compute":"cuda_core","numerics":"reference","where":"device","state":"conv:update"},{"id":"l25.gdn.conv_ckpt","order":593,"layer":25,"block":"l25.gdn","block_type":"gdn","kind":"linear_attention","op":"state_snapshot","roles":[],"inserted":false,"precision":{"weight":null,"activation":"f32","output":null},"steps":["state f32"],"group":359,"kernel":"causal_conv1d::causal_conv1d_update_l2norm_f32","family":"causal_conv1d_l2norm","point":null,"compute":"cuda_core","numerics":"reference","where":"device","state":"conv:snapshot"},{"id":"l25.gdn.l2","order":594,"layer":25,"block":"l25.gdn","block_type":"gdn","kind":"linear_attention","op":"l2_norm","roles":[],"inserted":false,"precision":{"weight":null,"activation":"f32","output":"f32"},"steps":["compute f32"],"group":359,"kernel":"causal_conv1d::causal_conv1d_update_l2norm_f32","family":"causal_conv1d_l2norm","point":null,"compute":"cuda_core","numerics":"reference","where":"device","state":null},{"id":"l25.gdn.recur","order":595,"layer":25,"block":"l25.gdn","block_type":"gdn","kind":"linear_attention","op":"gdn_recurrence","roles":[],"inserted":false,"precision":{"weight":null,"activation":"f32","output":"f32"},"steps":["state f32","compute f32"],"group":360,"kernel":"gated_delta_rule::gated_delta_rule_decode_f32","family":"gdn_recurrence","point":{"decay":"head","gate_act":"silu"},"compute":"cuda_core","numerics":"reference","where":"device","state":"h:update"},{"id":"l25.gdn.out_norm","order":596,"layer":25,"block":"l25.gdn","block_type":"gdn","kind":"linear_attention","op":"gated_rms_norm","roles":[],"inserted":false,"precision":{"weight":null,"activation":"f32","output":"bf16"},"steps":["compute f32"],"group":361,"kernel":"norm::gated_rms_norm_f32_input","family":"gated_rms_norm","point":null,"compute":"memory","numerics":"reference","where":"device","state":null},{"id":"l25.gdn.out","order":597,"layer":25,"block":"l25.gdn","block_type":"gdn","kind":"linear_attention","op":"linear:gdn_out","roles":["gdn_out"],"inserted":false,"precision":{"weight":"fp8/block128x128","activation":"bf16","output":"bf16"},"steps":["act bf16","weight fp8/block128x128->bf16","mma bf16*bf16","accumulate f32","scale f32"],"group":362,"kernel":"w8a16_gemm_pipelined_m32::w8a16_gemm_pipelined_m32","family":"w8a16_gemm","point":null,"compute":"tensor_core:mma.sync.m16n8k16.bf16","numerics":"reference","where":"device","state":null},{"id":"l25.gdn.add","order":598,"layer":25,"block":"l25.gdn","block_type":"gdn","kind":"linear_attention","op":"residual_add","roles":[],"inserted":false,"precision":{"weight":null,"activation":"bf16","output":"bf16"},"steps":["compute f32"],"group":363,"kernel":"norm::residual_add_rms_norm","family":"rms_norm","point":null,"compute":"memory","numerics":"reference","where":"device","state":null},{"id":"l25.moe_ffn.post_norm","order":599,"layer":25,"block":"l25.moe_ffn","block_type":"moe_ffn","kind":"linear_attention","op":"rms_norm","roles":[],"inserted":false,"precision":{"weight":null,"activation":"f32","output":"bf16"},"steps":["compute f32"],"group":363,"kernel":"norm::residual_add_rms_norm","family":"rms_norm","point":null,"compute":"memory","numerics":"reference","where":"device","state":null},{"id":"l25.moe_ffn.router","order":600,"layer":25,"block":"l25.moe_ffn","block_type":"moe_ffn","kind":"linear_attention","op":"router","roles":[],"inserted":false,"precision":{"weight":"bf16","activation":"bf16","output":"bf16"},"steps":["act bf16","weight bf16->bf16","mma bf16*bf16","accumulate f32","scale none"],"group":364,"kernel":"gemv::dense_gemv_bf16","family":"dense_bf16","point":null,"compute":"cuda_core","numerics":"reference","where":"device","state":null},{"id":"l25.moe_ffn.top_k","order":601,"layer":25,"block":"l25.moe_ffn","block_type":"moe_ffn","kind":"linear_attention","op":"top_k","roles":[],"inserted":false,"precision":{"weight":null,"activation":"bf16","output":"f32,i32"},"steps":["score f32"],"group":365,"kernel":"moe_topk::moe_topk_softmax","family":"moe_topk","point":null,"compute":"memory","numerics":"reference","where":"device","state":null},{"id":"l25.moe_ffn.experts_gate_up","order":602,"layer":25,"block":"l25.moe_ffn","block_type":"moe_ffn","kind":"linear_attention","op":"expert_gate_up","roles":[],"inserted":false,"precision":{"weight":"fp8/block128x128","activation":"bf16","output":"bf16"},"steps":["gather bf16","act bf16","weight fp8/block128x128->bf16","mma bf16*bf16","accumulate f32","scale f32"],"group":366,"kernel":"moe_shared_expert_fused_fp8::moe_expert_gate_up_shared_fp8","family":"moe_fp8_1row","point":null,"compute":"cuda_core","numerics":"reference","where":"device","state":null},{"id":"l25.moe_ffn.shared_gate_up","order":603,"layer":25,"block":"l25.moe_ffn","block_type":"moe_ffn","kind":"linear_attention","op":"linear:shared_gate_up","roles":["shared_gate_up"],"inserted":false,"precision":{"weight":"fp8/block128x128","activation":"bf16","output":"bf16"},"steps":["act bf16","weight fp8/block128x128->bf16","mma bf16*bf16","accumulate f32","scale f32"],"group":366,"kernel":"moe_shared_expert_fused_fp8::moe_expert_gate_up_shared_fp8","family":"moe_fp8_1row","point":null,"compute":"cuda_core","numerics":"reference","where":"device","state":null},{"id":"l25.moe_ffn.experts_act","order":604,"layer":25,"block":"l25.moe_ffn","block_type":"moe_ffn","kind":"linear_attention","op":"silu_mul","roles":[],"inserted":false,"precision":{"weight":null,"activation":"bf16","output":"f32"},"steps":["compute f32"],"group":367,"kernel":"moe_shared_expert_fused_fp8::moe_expert_silu_down_shared_fp8","family":"moe_fp8_1row","point":null,"compute":"cuda_core","numerics":"reference","where":"device","state":null},{"id":"l25.moe_ffn.experts_down","order":605,"layer":25,"block":"l25.moe_ffn","block_type":"moe_ffn","kind":"linear_attention","op":"expert_down","roles":[],"inserted":false,"precision":{"weight":"fp8/block128x128","activation":"f32","output":"bf16"},"steps":["act f32","weight fp8/block128x128->f32","mma f32*f32","accumulate f32","scale f32"],"group":367,"kernel":"moe_shared_expert_fused_fp8::moe_expert_silu_down_shared_fp8","family":"moe_fp8_1row","point":null,"compute":"cuda_core","numerics":"reference","where":"device","state":null},{"id":"l25.moe_ffn.shared_act","order":606,"layer":25,"block":"l25.moe_ffn","block_type":"moe_ffn","kind":"linear_attention","op":"silu_mul","roles":[],"inserted":false,"precision":{"weight":null,"activation":"bf16","output":"f32"},"steps":["compute f32"],"group":367,"kernel":"moe_shared_expert_fused_fp8::moe_expert_silu_down_shared_fp8","family":"moe_fp8_1row","point":null,"compute":"cuda_core","numerics":"reference","where":"device","state":null},{"id":"l25.moe_ffn.shared_down","order":607,"layer":25,"block":"l25.moe_ffn","block_type":"moe_ffn","kind":"linear_attention","op":"linear:shared_down","roles":["shared_down"],"inserted":false,"precision":{"weight":"fp8/block128x128","activation":"f32","output":"bf16"},"steps":["act f32","weight fp8/block128x128->f32","mma f32*f32","accumulate f32","scale f32"],"group":367,"kernel":"moe_shared_expert_fused_fp8::moe_expert_silu_down_shared_fp8","family":"moe_fp8_1row","point":null,"compute":"cuda_core","numerics":"reference","where":"device","state":null},{"id":"l25.moe_ffn.shared_gate","order":608,"layer":25,"block":"l25.moe_ffn","block_type":"moe_ffn","kind":"linear_attention","op":"linear:shared_gate","roles":["shared_gate"],"inserted":false,"precision":{"weight":"bf16","activation":"bf16","output":"f32"},"steps":["act bf16","weight bf16->bf16","mma bf16*bf16","accumulate f32","scale none"],"group":368,"kernel":"moe_expert_gemv::moe_weighted_sum_blend","family":"moe_blend","point":{"shared_gate":"sigmoid"},"compute":"cuda_core","numerics":"reference","where":"device","state":null},{"id":"l25.moe_ffn.blend","order":609,"layer":25,"block":"l25.moe_ffn","block_type":"moe_ffn","kind":"linear_attention","op":"blend","roles":[],"inserted":false,"precision":{"weight":null,"activation":"bf16","output":"bf16"},"steps":["scatter f32","combine f32"],"group":368,"kernel":"moe_expert_gemv::moe_weighted_sum_blend","family":"moe_blend","point":{"shared_gate":"sigmoid"},"compute":"cuda_core","numerics":"reference","where":"device","state":null},{"id":"l25.moe_ffn.add","order":610,"layer":25,"block":"l25.moe_ffn","block_type":"moe_ffn","kind":"linear_attention","op":"residual_add","roles":[],"inserted":false,"precision":{"weight":null,"activation":"bf16","output":"bf16"},"steps":["compute f32"],"group":369,"kernel":"residual_add_rms_norm_exact::residual_add_rms_norm_exact","family":"rms_norm","point":null,"compute":"memory","numerics":"bit_identical(residual_add_rms_norm_exact_microtest)","where":"device","state":null},{"id":"l26.gdn.input_norm","order":611,"layer":26,"block":"l26.gdn","block_type":"gdn","kind":"linear_attention","op":"rms_norm","roles":[],"inserted":false,"precision":{"weight":null,"activation":"bf16","output":"bf16"},"steps":["compute f32"],"group":369,"kernel":"residual_add_rms_norm_exact::residual_add_rms_norm_exact","family":"rms_norm","point":null,"compute":"memory","numerics":"bit_identical(residual_add_rms_norm_exact_microtest)","where":"device","state":null},{"id":"l26.gdn.qkvz","order":612,"layer":26,"block":"l26.gdn","block_type":"gdn","kind":"linear_attention","op":"linear:qkvz","roles":["qkvz"],"inserted":false,"precision":{"weight":"fp8/block128x128","activation":"bf16","output":"bf16"},"steps":["act bf16","weight fp8/block128x128->bf16","mma bf16*bf16","accumulate f32","scale f32"],"group":370,"kernel":"w8a16_gemm_pipelined_m32::w8a16_gemm_pipelined_m32","family":"w8a16_gemm","point":null,"compute":"tensor_core:mma.sync.m16n8k16.bf16","numerics":"reference","where":"device","state":null},{"id":"l26.gdn.ba","order":613,"layer":26,"block":"l26.gdn","block_type":"gdn","kind":"linear_attention","op":"linear:ba","roles":["ba"],"inserted":false,"precision":{"weight":"bf16","activation":"bf16","output":"f32"},"steps":["act bf16","weight bf16->bf16","mma bf16*bf16","accumulate f32","scale none"],"group":371,"kernel":"ssm_preprocess::dense_gemv_ba_gates","family":"gdn_ba_gates_gemv","point":{"gate":"softplus","decay":"head"},"compute":"cuda_core","numerics":"reference","where":"device","state":null},{"id":"l26.gdn.gates","order":614,"layer":26,"block":"l26.gdn","block_type":"gdn","kind":"linear_attention","op":"gdn_gates","roles":[],"inserted":false,"precision":{"weight":null,"activation":"f32","output":"f32,f32"},"steps":["compute f32"],"group":371,"kernel":"ssm_preprocess::dense_gemv_ba_gates","family":"gdn_ba_gates_gemv","point":{"gate":"softplus","decay":"head"},"compute":"cuda_core","numerics":"reference","where":"device","state":null},{"id":"l26.gdn.conv","order":615,"layer":26,"block":"l26.gdn","block_type":"gdn","kind":"linear_attention","op":"conv1d_update","roles":[],"inserted":false,"precision":{"weight":null,"activation":"bf16","output":"f32"},"steps":["state f32","compute f32"],"group":372,"kernel":"causal_conv1d::causal_conv1d_update_l2norm_f32","family":"causal_conv1d_l2norm","point":null,"compute":"cuda_core","numerics":"reference","where":"device","state":"conv:update"},{"id":"l26.gdn.conv_ckpt","order":616,"layer":26,"block":"l26.gdn","block_type":"gdn","kind":"linear_attention","op":"state_snapshot","roles":[],"inserted":false,"precision":{"weight":null,"activation":"f32","output":null},"steps":["state f32"],"group":372,"kernel":"causal_conv1d::causal_conv1d_update_l2norm_f32","family":"causal_conv1d_l2norm","point":null,"compute":"cuda_core","numerics":"reference","where":"device","state":"conv:snapshot"},{"id":"l26.gdn.l2","order":617,"layer":26,"block":"l26.gdn","block_type":"gdn","kind":"linear_attention","op":"l2_norm","roles":[],"inserted":false,"precision":{"weight":null,"activation":"f32","output":"f32"},"steps":["compute f32"],"group":372,"kernel":"causal_conv1d::causal_conv1d_update_l2norm_f32","family":"causal_conv1d_l2norm","point":null,"compute":"cuda_core","numerics":"reference","where":"device","state":null},{"id":"l26.gdn.recur","order":618,"layer":26,"block":"l26.gdn","block_type":"gdn","kind":"linear_attention","op":"gdn_recurrence","roles":[],"inserted":false,"precision":{"weight":null,"activation":"f32","output":"f32"},"steps":["state f32","compute f32"],"group":373,"kernel":"gated_delta_rule::gated_delta_rule_decode_f32","family":"gdn_recurrence","point":{"decay":"head","gate_act":"silu"},"compute":"cuda_core","numerics":"reference","where":"device","state":"h:update"},{"id":"l26.gdn.out_norm","order":619,"layer":26,"block":"l26.gdn","block_type":"gdn","kind":"linear_attention","op":"gated_rms_norm","roles":[],"inserted":false,"precision":{"weight":null,"activation":"f32","output":"bf16"},"steps":["compute f32"],"group":374,"kernel":"norm::gated_rms_norm_f32_input","family":"gated_rms_norm","point":null,"compute":"memory","numerics":"reference","where":"device","state":null},{"id":"l26.gdn.out","order":620,"layer":26,"block":"l26.gdn","block_type":"gdn","kind":"linear_attention","op":"linear:gdn_out","roles":["gdn_out"],"inserted":false,"precision":{"weight":"fp8/block128x128","activation":"bf16","output":"bf16"},"steps":["act bf16","weight fp8/block128x128->bf16","mma bf16*bf16","accumulate f32","scale f32"],"group":375,"kernel":"w8a16_gemm_pipelined_m32::w8a16_gemm_pipelined_m32","family":"w8a16_gemm","point":null,"compute":"tensor_core:mma.sync.m16n8k16.bf16","numerics":"reference","where":"device","state":null},{"id":"l26.gdn.add","order":621,"layer":26,"block":"l26.gdn","block_type":"gdn","kind":"linear_attention","op":"residual_add","roles":[],"inserted":false,"precision":{"weight":null,"activation":"bf16","output":"bf16"},"steps":["compute f32"],"group":376,"kernel":"norm::residual_add_rms_norm","family":"rms_norm","point":null,"compute":"memory","numerics":"reference","where":"device","state":null},{"id":"l26.moe_ffn.post_norm","order":622,"layer":26,"block":"l26.moe_ffn","block_type":"moe_ffn","kind":"linear_attention","op":"rms_norm","roles":[],"inserted":false,"precision":{"weight":null,"activation":"f32","output":"bf16"},"steps":["compute f32"],"group":376,"kernel":"norm::residual_add_rms_norm","family":"rms_norm","point":null,"compute":"memory","numerics":"reference","where":"device","state":null},{"id":"l26.moe_ffn.router","order":623,"layer":26,"block":"l26.moe_ffn","block_type":"moe_ffn","kind":"linear_attention","op":"router","roles":[],"inserted":false,"precision":{"weight":"bf16","activation":"bf16","output":"bf16"},"steps":["act bf16","weight bf16->bf16","mma bf16*bf16","accumulate f32","scale none"],"group":377,"kernel":"gemv::dense_gemv_bf16","family":"dense_bf16","point":null,"compute":"cuda_core","numerics":"reference","where":"device","state":null},{"id":"l26.moe_ffn.top_k","order":624,"layer":26,"block":"l26.moe_ffn","block_type":"moe_ffn","kind":"linear_attention","op":"top_k","roles":[],"inserted":false,"precision":{"weight":null,"activation":"bf16","output":"f32,i32"},"steps":["score f32"],"group":378,"kernel":"moe_topk::moe_topk_softmax","family":"moe_topk","point":null,"compute":"memory","numerics":"reference","where":"device","state":null},{"id":"l26.moe_ffn.experts_gate_up","order":625,"layer":26,"block":"l26.moe_ffn","block_type":"moe_ffn","kind":"linear_attention","op":"expert_gate_up","roles":[],"inserted":false,"precision":{"weight":"fp8/block128x128","activation":"bf16","output":"bf16"},"steps":["gather bf16","act bf16","weight fp8/block128x128->bf16","mma bf16*bf16","accumulate f32","scale f32"],"group":379,"kernel":"moe_shared_expert_fused_fp8::moe_expert_gate_up_shared_fp8","family":"moe_fp8_1row","point":null,"compute":"cuda_core","numerics":"reference","where":"device","state":null},{"id":"l26.moe_ffn.shared_gate_up","order":626,"layer":26,"block":"l26.moe_ffn","block_type":"moe_ffn","kind":"linear_attention","op":"linear:shared_gate_up","roles":["shared_gate_up"],"inserted":false,"precision":{"weight":"fp8/block128x128","activation":"bf16","output":"bf16"},"steps":["act bf16","weight fp8/block128x128->bf16","mma bf16*bf16","accumulate f32","scale f32"],"group":379,"kernel":"moe_shared_expert_fused_fp8::moe_expert_gate_up_shared_fp8","family":"moe_fp8_1row","point":null,"compute":"cuda_core","numerics":"reference","where":"device","state":null},{"id":"l26.moe_ffn.experts_act","order":627,"layer":26,"block":"l26.moe_ffn","block_type":"moe_ffn","kind":"linear_attention","op":"silu_mul","roles":[],"inserted":false,"precision":{"weight":null,"activation":"bf16","output":"f32"},"steps":["compute f32"],"group":380,"kernel":"moe_shared_expert_fused_fp8::moe_expert_silu_down_shared_fp8","family":"moe_fp8_1row","point":null,"compute":"cuda_core","numerics":"reference","where":"device","state":null},{"id":"l26.moe_ffn.experts_down","order":628,"layer":26,"block":"l26.moe_ffn","block_type":"moe_ffn","kind":"linear_attention","op":"expert_down","roles":[],"inserted":false,"precision":{"weight":"fp8/block128x128","activation":"f32","output":"bf16"},"steps":["act f32","weight fp8/block128x128->f32","mma f32*f32","accumulate f32","scale f32"],"group":380,"kernel":"moe_shared_expert_fused_fp8::moe_expert_silu_down_shared_fp8","family":"moe_fp8_1row","point":null,"compute":"cuda_core","numerics":"reference","where":"device","state":null},{"id":"l26.moe_ffn.shared_act","order":629,"layer":26,"block":"l26.moe_ffn","block_type":"moe_ffn","kind":"linear_attention","op":"silu_mul","roles":[],"inserted":false,"precision":{"weight":null,"activation":"bf16","output":"f32"},"steps":["compute f32"],"group":380,"kernel":"moe_shared_expert_fused_fp8::moe_expert_silu_down_shared_fp8","family":"moe_fp8_1row","point":null,"compute":"cuda_core","numerics":"reference","where":"device","state":null},{"id":"l26.moe_ffn.shared_down","order":630,"layer":26,"block":"l26.moe_ffn","block_type":"moe_ffn","kind":"linear_attention","op":"linear:shared_down","roles":["shared_down"],"inserted":false,"precision":{"weight":"fp8/block128x128","activation":"f32","output":"bf16"},"steps":["act f32","weight fp8/block128x128->f32","mma f32*f32","accumulate f32","scale f32"],"group":380,"kernel":"moe_shared_expert_fused_fp8::moe_expert_silu_down_shared_fp8","family":"moe_fp8_1row","point":null,"compute":"cuda_core","numerics":"reference","where":"device","state":null},{"id":"l26.moe_ffn.shared_gate","order":631,"layer":26,"block":"l26.moe_ffn","block_type":"moe_ffn","kind":"linear_attention","op":"linear:shared_gate","roles":["shared_gate"],"inserted":false,"precision":{"weight":"bf16","activation":"bf16","output":"f32"},"steps":["act bf16","weight bf16->bf16","mma bf16*bf16","accumulate f32","scale none"],"group":381,"kernel":"moe_expert_gemv::moe_weighted_sum_blend","family":"moe_blend","point":{"shared_gate":"sigmoid"},"compute":"cuda_core","numerics":"reference","where":"device","state":null},{"id":"l26.moe_ffn.blend","order":632,"layer":26,"block":"l26.moe_ffn","block_type":"moe_ffn","kind":"linear_attention","op":"blend","roles":[],"inserted":false,"precision":{"weight":null,"activation":"bf16","output":"bf16"},"steps":["scatter f32","combine f32"],"group":381,"kernel":"moe_expert_gemv::moe_weighted_sum_blend","family":"moe_blend","point":{"shared_gate":"sigmoid"},"compute":"cuda_core","numerics":"reference","where":"device","state":null},{"id":"l26.moe_ffn.add","order":633,"layer":26,"block":"l26.moe_ffn","block_type":"moe_ffn","kind":"linear_attention","op":"residual_add","roles":[],"inserted":false,"precision":{"weight":null,"activation":"bf16","output":"bf16"},"steps":["compute f32"],"group":382,"kernel":"residual_add_rms_norm_exact::residual_add_rms_norm_exact","family":"rms_norm","point":null,"compute":"memory","numerics":"bit_identical(residual_add_rms_norm_exact_microtest)","where":"device","state":null},{"id":"l27.attn.input_norm","order":634,"layer":27,"block":"l27.attn","block_type":"attn","kind":"full_attention","op":"rms_norm","roles":[],"inserted":false,"precision":{"weight":null,"activation":"bf16","output":"bf16"},"steps":["compute f32"],"group":382,"kernel":"residual_add_rms_norm_exact::residual_add_rms_norm_exact","family":"rms_norm","point":null,"compute":"memory","numerics":"bit_identical(residual_add_rms_norm_exact_microtest)","where":"device","state":null},{"id":"l27.attn.q","order":635,"layer":27,"block":"l27.attn","block_type":"attn","kind":"full_attention","op":"linear:q","roles":["q"],"inserted":false,"precision":{"weight":"fp8/block128x128","activation":"bf16","output":"bf16"},"steps":["act bf16","weight fp8/block128x128->bf16","mma bf16*bf16","accumulate f32","scale f32"],"group":383,"kernel":"w8a16_gemm_pipelined_m32::w8a16_gemm_pipelined_m32","family":"w8a16_gemm","point":null,"compute":"tensor_core:mma.sync.m16n8k16.bf16","numerics":"reference","where":"device","state":null},{"id":"l27.attn.q_split","order":636,"layer":27,"block":"l27.attn","block_type":"attn","kind":"full_attention","op":"split","roles":[],"inserted":false,"precision":{"weight":null,"activation":"bf16","output":"bf16,bf16"},"steps":["move bf16"],"group":384,"kernel":"ssm_preprocess::deinterleave_qg","family":"deinterleave_qg","point":{"layout":"interleaved"},"compute":"memory","numerics":"reference","where":"device","state":null},{"id":"l27.attn.k","order":637,"layer":27,"block":"l27.attn","block_type":"attn","kind":"full_attention","op":"linear:k","roles":["k"],"inserted":false,"precision":{"weight":"fp8/block128x128","activation":"bf16","output":"bf16"},"steps":["act bf16","weight fp8/block128x128->bf16","mma bf16*bf16","accumulate f32","scale f32"],"group":385,"kernel":"w8a16_gemm_pipelined_m32::w8a16_gemm_pipelined_m32","family":"w8a16_gemm","point":null,"compute":"tensor_core:mma.sync.m16n8k16.bf16","numerics":"reference","where":"device","state":null},{"id":"l27.attn.v","order":638,"layer":27,"block":"l27.attn","block_type":"attn","kind":"full_attention","op":"linear:v","roles":["v"],"inserted":false,"precision":{"weight":"fp8/block128x128","activation":"bf16","output":"bf16"},"steps":["act bf16","weight fp8/block128x128->bf16","mma bf16*bf16","accumulate f32","scale f32"],"group":386,"kernel":"w8a16_gemm_pipelined_m32::w8a16_gemm_pipelined_m32","family":"w8a16_gemm","point":null,"compute":"tensor_core:mma.sync.m16n8k16.bf16","numerics":"reference","where":"device","state":null},{"id":"l27.attn.q_norm","order":639,"layer":27,"block":"l27.attn","block_type":"attn","kind":"full_attention","op":"qk_norm","roles":[],"inserted":false,"precision":{"weight":null,"activation":"bf16","output":"bf16"},"steps":["compute f32"],"group":387,"kernel":"norm::rms_norm","family":"rms_norm","point":null,"compute":"memory","numerics":"reference","where":"device","state":null},{"id":"l27.attn.k_norm","order":640,"layer":27,"block":"l27.attn","block_type":"attn","kind":"full_attention","op":"qk_norm","roles":[],"inserted":false,"precision":{"weight":null,"activation":"bf16","output":"bf16"},"steps":["compute f32"],"group":388,"kernel":"norm::rms_norm","family":"rms_norm","point":null,"compute":"memory","numerics":"reference","where":"device","state":null},{"id":"l27.attn.rope","order":641,"layer":27,"block":"l27.attn","block_type":"attn","kind":"full_attention","op":"rope","roles":[],"inserted":false,"precision":{"weight":null,"activation":"bf16","output":"bf16,bf16"},"steps":["compute f32"],"group":389,"kernel":"rope_mrope_interleaved::rope_forward_mrope_interleaved","family":"rope","point":null,"compute":"memory","numerics":"reference","where":"device","state":null},{"id":"l27.attn.kv_write","order":642,"layer":27,"block":"l27.attn","block_type":"attn","kind":"full_attention","op":"kv_write","roles":[],"inserted":false,"precision":{"weight":null,"activation":"bf16","output":null},"steps":["cache bf16"],"group":390,"kernel":"reshape_and_cache::reshape_and_cache_flash","family":"kv_write","point":null,"compute":"memory","numerics":"reference","where":"device","state":"k:write,v:write"},{"id":"l27.attn.attend","order":643,"layer":27,"block":"l27.attn","block_type":"attn","kind":"full_attention","op":"paged_attention","roles":[],"inserted":false,"precision":{"weight":null,"activation":"bf16","output":"bf16"},"steps":["cache bf16","scores f32","softmax f32","accumulate f32"],"group":391,"kernel":"paged_decode::paged_decode_attn","family":"paged_decode_attn","point":null,"compute":"cuda_core","numerics":"reference","where":"device","state":"k:read,v:read"},{"id":"l27.attn.gate_mul","order":644,"layer":27,"block":"l27.attn","block_type":"attn","kind":"full_attention","op":"sigmoid_gate_mul","roles":[],"inserted":false,"precision":{"weight":null,"activation":"bf16","output":"bf16"},"steps":["compute f32"],"group":392,"kernel":"residual_add::sigmoid_gate_mul","family":"sigmoid_gate_mul","point":null,"compute":"memory","numerics":"reference","where":"device","state":null},{"id":"l27.attn.o","order":645,"layer":27,"block":"l27.attn","block_type":"attn","kind":"full_attention","op":"linear:o","roles":["o"],"inserted":false,"precision":{"weight":"fp8/block128x128","activation":"bf16","output":"bf16"},"steps":["act bf16","weight fp8/block128x128->bf16","mma bf16*bf16","accumulate f32","scale f32"],"group":393,"kernel":"w8a16_gemm_pipelined_m32::w8a16_gemm_pipelined_m32","family":"w8a16_gemm","point":null,"compute":"tensor_core:mma.sync.m16n8k16.bf16","numerics":"reference","where":"device","state":null},{"id":"l27.attn.add","order":646,"layer":27,"block":"l27.attn","block_type":"attn","kind":"full_attention","op":"residual_add","roles":[],"inserted":false,"precision":{"weight":null,"activation":"bf16","output":"bf16"},"steps":["compute f32"],"group":394,"kernel":"norm::residual_add_rms_norm","family":"rms_norm","point":null,"compute":"memory","numerics":"reference","where":"device","state":null},{"id":"l27.moe_ffn.post_norm","order":647,"layer":27,"block":"l27.moe_ffn","block_type":"moe_ffn","kind":"full_attention","op":"rms_norm","roles":[],"inserted":false,"precision":{"weight":null,"activation":"f32","output":"bf16"},"steps":["compute f32"],"group":394,"kernel":"norm::residual_add_rms_norm","family":"rms_norm","point":null,"compute":"memory","numerics":"reference","where":"device","state":null},{"id":"l27.moe_ffn.router","order":648,"layer":27,"block":"l27.moe_ffn","block_type":"moe_ffn","kind":"full_attention","op":"router","roles":[],"inserted":false,"precision":{"weight":"bf16","activation":"bf16","output":"bf16"},"steps":["act bf16","weight bf16->bf16","mma bf16*bf16","accumulate f32","scale none"],"group":395,"kernel":"gemv::dense_gemv_bf16","family":"dense_bf16","point":null,"compute":"cuda_core","numerics":"reference","where":"device","state":null},{"id":"l27.moe_ffn.top_k","order":649,"layer":27,"block":"l27.moe_ffn","block_type":"moe_ffn","kind":"full_attention","op":"top_k","roles":[],"inserted":false,"precision":{"weight":null,"activation":"bf16","output":"f32,i32"},"steps":["score f32"],"group":396,"kernel":"moe_topk::moe_topk_softmax","family":"moe_topk","point":null,"compute":"memory","numerics":"reference","where":"device","state":null},{"id":"l27.moe_ffn.experts_gate_up","order":650,"layer":27,"block":"l27.moe_ffn","block_type":"moe_ffn","kind":"full_attention","op":"expert_gate_up","roles":[],"inserted":false,"precision":{"weight":"fp8/block128x128","activation":"bf16","output":"bf16"},"steps":["gather bf16","act bf16","weight fp8/block128x128->bf16","mma bf16*bf16","accumulate f32","scale f32"],"group":397,"kernel":"moe_shared_expert_fused_fp8::moe_expert_gate_up_shared_fp8","family":"moe_fp8_1row","point":null,"compute":"cuda_core","numerics":"reference","where":"device","state":null},{"id":"l27.moe_ffn.shared_gate_up","order":651,"layer":27,"block":"l27.moe_ffn","block_type":"moe_ffn","kind":"full_attention","op":"linear:shared_gate_up","roles":["shared_gate_up"],"inserted":false,"precision":{"weight":"fp8/block128x128","activation":"bf16","output":"bf16"},"steps":["act bf16","weight fp8/block128x128->bf16","mma bf16*bf16","accumulate f32","scale f32"],"group":397,"kernel":"moe_shared_expert_fused_fp8::moe_expert_gate_up_shared_fp8","family":"moe_fp8_1row","point":null,"compute":"cuda_core","numerics":"reference","where":"device","state":null},{"id":"l27.moe_ffn.experts_act","order":652,"layer":27,"block":"l27.moe_ffn","block_type":"moe_ffn","kind":"full_attention","op":"silu_mul","roles":[],"inserted":false,"precision":{"weight":null,"activation":"bf16","output":"f32"},"steps":["compute f32"],"group":398,"kernel":"moe_shared_expert_fused_fp8::moe_expert_silu_down_shared_fp8","family":"moe_fp8_1row","point":null,"compute":"cuda_core","numerics":"reference","where":"device","state":null},{"id":"l27.moe_ffn.experts_down","order":653,"layer":27,"block":"l27.moe_ffn","block_type":"moe_ffn","kind":"full_attention","op":"expert_down","roles":[],"inserted":false,"precision":{"weight":"fp8/block128x128","activation":"f32","output":"bf16"},"steps":["act f32","weight fp8/block128x128->f32","mma f32*f32","accumulate f32","scale f32"],"group":398,"kernel":"moe_shared_expert_fused_fp8::moe_expert_silu_down_shared_fp8","family":"moe_fp8_1row","point":null,"compute":"cuda_core","numerics":"reference","where":"device","state":null},{"id":"l27.moe_ffn.shared_act","order":654,"layer":27,"block":"l27.moe_ffn","block_type":"moe_ffn","kind":"full_attention","op":"silu_mul","roles":[],"inserted":false,"precision":{"weight":null,"activation":"bf16","output":"f32"},"steps":["compute f32"],"group":398,"kernel":"moe_shared_expert_fused_fp8::moe_expert_silu_down_shared_fp8","family":"moe_fp8_1row","point":null,"compute":"cuda_core","numerics":"reference","where":"device","state":null},{"id":"l27.moe_ffn.shared_down","order":655,"layer":27,"block":"l27.moe_ffn","block_type":"moe_ffn","kind":"full_attention","op":"linear:shared_down","roles":["shared_down"],"inserted":false,"precision":{"weight":"fp8/block128x128","activation":"f32","output":"bf16"},"steps":["act f32","weight fp8/block128x128->f32","mma f32*f32","accumulate f32","scale f32"],"group":398,"kernel":"moe_shared_expert_fused_fp8::moe_expert_silu_down_shared_fp8","family":"moe_fp8_1row","point":null,"compute":"cuda_core","numerics":"reference","where":"device","state":null},{"id":"l27.moe_ffn.shared_gate","order":656,"layer":27,"block":"l27.moe_ffn","block_type":"moe_ffn","kind":"full_attention","op":"linear:shared_gate","roles":["shared_gate"],"inserted":false,"precision":{"weight":"bf16","activation":"bf16","output":"f32"},"steps":["act bf16","weight bf16->bf16","mma bf16*bf16","accumulate f32","scale none"],"group":399,"kernel":"moe_expert_gemv::moe_weighted_sum_blend","family":"moe_blend","point":{"shared_gate":"sigmoid"},"compute":"cuda_core","numerics":"reference","where":"device","state":null},{"id":"l27.moe_ffn.blend","order":657,"layer":27,"block":"l27.moe_ffn","block_type":"moe_ffn","kind":"full_attention","op":"blend","roles":[],"inserted":false,"precision":{"weight":null,"activation":"bf16","output":"bf16"},"steps":["scatter f32","combine f32"],"group":399,"kernel":"moe_expert_gemv::moe_weighted_sum_blend","family":"moe_blend","point":{"shared_gate":"sigmoid"},"compute":"cuda_core","numerics":"reference","where":"device","state":null},{"id":"l27.moe_ffn.add","order":658,"layer":27,"block":"l27.moe_ffn","block_type":"moe_ffn","kind":"full_attention","op":"residual_add","roles":[],"inserted":false,"precision":{"weight":null,"activation":"bf16","output":"bf16"},"steps":["compute f32"],"group":400,"kernel":"residual_add_rms_norm_exact::residual_add_rms_norm_exact","family":"rms_norm","point":null,"compute":"memory","numerics":"bit_identical(residual_add_rms_norm_exact_microtest)","where":"device","state":null},{"id":"l28.gdn.input_norm","order":659,"layer":28,"block":"l28.gdn","block_type":"gdn","kind":"linear_attention","op":"rms_norm","roles":[],"inserted":false,"precision":{"weight":null,"activation":"bf16","output":"bf16"},"steps":["compute f32"],"group":400,"kernel":"residual_add_rms_norm_exact::residual_add_rms_norm_exact","family":"rms_norm","point":null,"compute":"memory","numerics":"bit_identical(residual_add_rms_norm_exact_microtest)","where":"device","state":null},{"id":"l28.gdn.qkvz","order":660,"layer":28,"block":"l28.gdn","block_type":"gdn","kind":"linear_attention","op":"linear:qkvz","roles":["qkvz"],"inserted":false,"precision":{"weight":"fp8/block128x128","activation":"bf16","output":"bf16"},"steps":["act bf16","weight fp8/block128x128->bf16","mma bf16*bf16","accumulate f32","scale f32"],"group":401,"kernel":"w8a16_gemm_pipelined_m32::w8a16_gemm_pipelined_m32","family":"w8a16_gemm","point":null,"compute":"tensor_core:mma.sync.m16n8k16.bf16","numerics":"reference","where":"device","state":null},{"id":"l28.gdn.ba","order":661,"layer":28,"block":"l28.gdn","block_type":"gdn","kind":"linear_attention","op":"linear:ba","roles":["ba"],"inserted":false,"precision":{"weight":"bf16","activation":"bf16","output":"f32"},"steps":["act bf16","weight bf16->bf16","mma bf16*bf16","accumulate f32","scale none"],"group":402,"kernel":"ssm_preprocess::dense_gemv_ba_gates","family":"gdn_ba_gates_gemv","point":{"gate":"softplus","decay":"head"},"compute":"cuda_core","numerics":"reference","where":"device","state":null},{"id":"l28.gdn.gates","order":662,"layer":28,"block":"l28.gdn","block_type":"gdn","kind":"linear_attention","op":"gdn_gates","roles":[],"inserted":false,"precision":{"weight":null,"activation":"f32","output":"f32,f32"},"steps":["compute f32"],"group":402,"kernel":"ssm_preprocess::dense_gemv_ba_gates","family":"gdn_ba_gates_gemv","point":{"gate":"softplus","decay":"head"},"compute":"cuda_core","numerics":"reference","where":"device","state":null},{"id":"l28.gdn.conv","order":663,"layer":28,"block":"l28.gdn","block_type":"gdn","kind":"linear_attention","op":"conv1d_update","roles":[],"inserted":false,"precision":{"weight":null,"activation":"bf16","output":"f32"},"steps":["state f32","compute f32"],"group":403,"kernel":"causal_conv1d::causal_conv1d_update_l2norm_f32","family":"causal_conv1d_l2norm","point":null,"compute":"cuda_core","numerics":"reference","where":"device","state":"conv:update"},{"id":"l28.gdn.conv_ckpt","order":664,"layer":28,"block":"l28.gdn","block_type":"gdn","kind":"linear_attention","op":"state_snapshot","roles":[],"inserted":false,"precision":{"weight":null,"activation":"f32","output":null},"steps":["state f32"],"group":403,"kernel":"causal_conv1d::causal_conv1d_update_l2norm_f32","family":"causal_conv1d_l2norm","point":null,"compute":"cuda_core","numerics":"reference","where":"device","state":"conv:snapshot"},{"id":"l28.gdn.l2","order":665,"layer":28,"block":"l28.gdn","block_type":"gdn","kind":"linear_attention","op":"l2_norm","roles":[],"inserted":false,"precision":{"weight":null,"activation":"f32","output":"f32"},"steps":["compute f32"],"group":403,"kernel":"causal_conv1d::causal_conv1d_update_l2norm_f32","family":"causal_conv1d_l2norm","point":null,"compute":"cuda_core","numerics":"reference","where":"device","state":null},{"id":"l28.gdn.recur","order":666,"layer":28,"block":"l28.gdn","block_type":"gdn","kind":"linear_attention","op":"gdn_recurrence","roles":[],"inserted":false,"precision":{"weight":null,"activation":"f32","output":"f32"},"steps":["state f32","compute f32"],"group":404,"kernel":"gated_delta_rule::gated_delta_rule_decode_f32","family":"gdn_recurrence","point":{"decay":"head","gate_act":"silu"},"compute":"cuda_core","numerics":"reference","where":"device","state":"h:update"},{"id":"l28.gdn.out_norm","order":667,"layer":28,"block":"l28.gdn","block_type":"gdn","kind":"linear_attention","op":"gated_rms_norm","roles":[],"inserted":false,"precision":{"weight":null,"activation":"f32","output":"bf16"},"steps":["compute f32"],"group":405,"kernel":"norm::gated_rms_norm_f32_input","family":"gated_rms_norm","point":null,"compute":"memory","numerics":"reference","where":"device","state":null},{"id":"l28.gdn.out","order":668,"layer":28,"block":"l28.gdn","block_type":"gdn","kind":"linear_attention","op":"linear:gdn_out","roles":["gdn_out"],"inserted":false,"precision":{"weight":"fp8/block128x128","activation":"bf16","output":"bf16"},"steps":["act bf16","weight fp8/block128x128->bf16","mma bf16*bf16","accumulate f32","scale f32"],"group":406,"kernel":"w8a16_gemm_pipelined_m32::w8a16_gemm_pipelined_m32","family":"w8a16_gemm","point":null,"compute":"tensor_core:mma.sync.m16n8k16.bf16","numerics":"reference","where":"device","state":null},{"id":"l28.gdn.add","order":669,"layer":28,"block":"l28.gdn","block_type":"gdn","kind":"linear_attention","op":"residual_add","roles":[],"inserted":false,"precision":{"weight":null,"activation":"bf16","output":"bf16"},"steps":["compute f32"],"group":407,"kernel":"norm::residual_add_rms_norm","family":"rms_norm","point":null,"compute":"memory","numerics":"reference","where":"device","state":null},{"id":"l28.moe_ffn.post_norm","order":670,"layer":28,"block":"l28.moe_ffn","block_type":"moe_ffn","kind":"linear_attention","op":"rms_norm","roles":[],"inserted":false,"precision":{"weight":null,"activation":"f32","output":"bf16"},"steps":["compute f32"],"group":407,"kernel":"norm::residual_add_rms_norm","family":"rms_norm","point":null,"compute":"memory","numerics":"reference","where":"device","state":null},{"id":"l28.moe_ffn.router","order":671,"layer":28,"block":"l28.moe_ffn","block_type":"moe_ffn","kind":"linear_attention","op":"router","roles":[],"inserted":false,"precision":{"weight":"bf16","activation":"bf16","output":"bf16"},"steps":["act bf16","weight bf16->bf16","mma bf16*bf16","accumulate f32","scale none"],"group":408,"kernel":"gemv::dense_gemv_bf16","family":"dense_bf16","point":null,"compute":"cuda_core","numerics":"reference","where":"device","state":null},{"id":"l28.moe_ffn.top_k","order":672,"layer":28,"block":"l28.moe_ffn","block_type":"moe_ffn","kind":"linear_attention","op":"top_k","roles":[],"inserted":false,"precision":{"weight":null,"activation":"bf16","output":"f32,i32"},"steps":["score f32"],"group":409,"kernel":"moe_topk::moe_topk_softmax","family":"moe_topk","point":null,"compute":"memory","numerics":"reference","where":"device","state":null},{"id":"l28.moe_ffn.experts_gate_up","order":673,"layer":28,"block":"l28.moe_ffn","block_type":"moe_ffn","kind":"linear_attention","op":"expert_gate_up","roles":[],"inserted":false,"precision":{"weight":"fp8/block128x128","activation":"bf16","output":"bf16"},"steps":["gather bf16","act bf16","weight fp8/block128x128->bf16","mma bf16*bf16","accumulate f32","scale f32"],"group":410,"kernel":"moe_shared_expert_fused_fp8::moe_expert_gate_up_shared_fp8","family":"moe_fp8_1row","point":null,"compute":"cuda_core","numerics":"reference","where":"device","state":null},{"id":"l28.moe_ffn.shared_gate_up","order":674,"layer":28,"block":"l28.moe_ffn","block_type":"moe_ffn","kind":"linear_attention","op":"linear:shared_gate_up","roles":["shared_gate_up"],"inserted":false,"precision":{"weight":"fp8/block128x128","activation":"bf16","output":"bf16"},"steps":["act bf16","weight fp8/block128x128->bf16","mma bf16*bf16","accumulate f32","scale f32"],"group":410,"kernel":"moe_shared_expert_fused_fp8::moe_expert_gate_up_shared_fp8","family":"moe_fp8_1row","point":null,"compute":"cuda_core","numerics":"reference","where":"device","state":null},{"id":"l28.moe_ffn.experts_act","order":675,"layer":28,"block":"l28.moe_ffn","block_type":"moe_ffn","kind":"linear_attention","op":"silu_mul","roles":[],"inserted":false,"precision":{"weight":null,"activation":"bf16","output":"f32"},"steps":["compute f32"],"group":411,"kernel":"moe_shared_expert_fused_fp8::moe_expert_silu_down_shared_fp8","family":"moe_fp8_1row","point":null,"compute":"cuda_core","numerics":"reference","where":"device","state":null},{"id":"l28.moe_ffn.experts_down","order":676,"layer":28,"block":"l28.moe_ffn","block_type":"moe_ffn","kind":"linear_attention","op":"expert_down","roles":[],"inserted":false,"precision":{"weight":"fp8/block128x128","activation":"f32","output":"bf16"},"steps":["act f32","weight fp8/block128x128->f32","mma f32*f32","accumulate f32","scale f32"],"group":411,"kernel":"moe_shared_expert_fused_fp8::moe_expert_silu_down_shared_fp8","family":"moe_fp8_1row","point":null,"compute":"cuda_core","numerics":"reference","where":"device","state":null},{"id":"l28.moe_ffn.shared_act","order":677,"layer":28,"block":"l28.moe_ffn","block_type":"moe_ffn","kind":"linear_attention","op":"silu_mul","roles":[],"inserted":false,"precision":{"weight":null,"activation":"bf16","output":"f32"},"steps":["compute f32"],"group":411,"kernel":"moe_shared_expert_fused_fp8::moe_expert_silu_down_shared_fp8","family":"moe_fp8_1row","point":null,"compute":"cuda_core","numerics":"reference","where":"device","state":null},{"id":"l28.moe_ffn.shared_down","order":678,"layer":28,"block":"l28.moe_ffn","block_type":"moe_ffn","kind":"linear_attention","op":"linear:shared_down","roles":["shared_down"],"inserted":false,"precision":{"weight":"fp8/block128x128","activation":"f32","output":"bf16"},"steps":["act f32","weight fp8/block128x128->f32","mma f32*f32","accumulate f32","scale f32"],"group":411,"kernel":"moe_shared_expert_fused_fp8::moe_expert_silu_down_shared_fp8","family":"moe_fp8_1row","point":null,"compute":"cuda_core","numerics":"reference","where":"device","state":null},{"id":"l28.moe_ffn.shared_gate","order":679,"layer":28,"block":"l28.moe_ffn","block_type":"moe_ffn","kind":"linear_attention","op":"linear:shared_gate","roles":["shared_gate"],"inserted":false,"precision":{"weight":"bf16","activation":"bf16","output":"f32"},"steps":["act bf16","weight bf16->bf16","mma bf16*bf16","accumulate f32","scale none"],"group":412,"kernel":"moe_expert_gemv::moe_weighted_sum_blend","family":"moe_blend","point":{"shared_gate":"sigmoid"},"compute":"cuda_core","numerics":"reference","where":"device","state":null},{"id":"l28.moe_ffn.blend","order":680,"layer":28,"block":"l28.moe_ffn","block_type":"moe_ffn","kind":"linear_attention","op":"blend","roles":[],"inserted":false,"precision":{"weight":null,"activation":"bf16","output":"bf16"},"steps":["scatter f32","combine f32"],"group":412,"kernel":"moe_expert_gemv::moe_weighted_sum_blend","family":"moe_blend","point":{"shared_gate":"sigmoid"},"compute":"cuda_core","numerics":"reference","where":"device","state":null},{"id":"l28.moe_ffn.add","order":681,"layer":28,"block":"l28.moe_ffn","block_type":"moe_ffn","kind":"linear_attention","op":"residual_add","roles":[],"inserted":false,"precision":{"weight":null,"activation":"bf16","output":"bf16"},"steps":["compute f32"],"group":413,"kernel":"residual_add_rms_norm_exact::residual_add_rms_norm_exact","family":"rms_norm","point":null,"compute":"memory","numerics":"bit_identical(residual_add_rms_norm_exact_microtest)","where":"device","state":null},{"id":"l29.gdn.input_norm","order":682,"layer":29,"block":"l29.gdn","block_type":"gdn","kind":"linear_attention","op":"rms_norm","roles":[],"inserted":false,"precision":{"weight":null,"activation":"bf16","output":"bf16"},"steps":["compute f32"],"group":413,"kernel":"residual_add_rms_norm_exact::residual_add_rms_norm_exact","family":"rms_norm","point":null,"compute":"memory","numerics":"bit_identical(residual_add_rms_norm_exact_microtest)","where":"device","state":null},{"id":"l29.gdn.qkvz","order":683,"layer":29,"block":"l29.gdn","block_type":"gdn","kind":"linear_attention","op":"linear:qkvz","roles":["qkvz"],"inserted":false,"precision":{"weight":"fp8/block128x128","activation":"bf16","output":"bf16"},"steps":["act bf16","weight fp8/block128x128->bf16","mma bf16*bf16","accumulate f32","scale f32"],"group":414,"kernel":"w8a16_gemm_pipelined_m32::w8a16_gemm_pipelined_m32","family":"w8a16_gemm","point":null,"compute":"tensor_core:mma.sync.m16n8k16.bf16","numerics":"reference","where":"device","state":null},{"id":"l29.gdn.ba","order":684,"layer":29,"block":"l29.gdn","block_type":"gdn","kind":"linear_attention","op":"linear:ba","roles":["ba"],"inserted":false,"precision":{"weight":"bf16","activation":"bf16","output":"f32"},"steps":["act bf16","weight bf16->bf16","mma bf16*bf16","accumulate f32","scale none"],"group":415,"kernel":"ssm_preprocess::dense_gemv_ba_gates","family":"gdn_ba_gates_gemv","point":{"gate":"softplus","decay":"head"},"compute":"cuda_core","numerics":"reference","where":"device","state":null},{"id":"l29.gdn.gates","order":685,"layer":29,"block":"l29.gdn","block_type":"gdn","kind":"linear_attention","op":"gdn_gates","roles":[],"inserted":false,"precision":{"weight":null,"activation":"f32","output":"f32,f32"},"steps":["compute f32"],"group":415,"kernel":"ssm_preprocess::dense_gemv_ba_gates","family":"gdn_ba_gates_gemv","point":{"gate":"softplus","decay":"head"},"compute":"cuda_core","numerics":"reference","where":"device","state":null},{"id":"l29.gdn.conv","order":686,"layer":29,"block":"l29.gdn","block_type":"gdn","kind":"linear_attention","op":"conv1d_update","roles":[],"inserted":false,"precision":{"weight":null,"activation":"bf16","output":"f32"},"steps":["state f32","compute f32"],"group":416,"kernel":"causal_conv1d::causal_conv1d_update_l2norm_f32","family":"causal_conv1d_l2norm","point":null,"compute":"cuda_core","numerics":"reference","where":"device","state":"conv:update"},{"id":"l29.gdn.conv_ckpt","order":687,"layer":29,"block":"l29.gdn","block_type":"gdn","kind":"linear_attention","op":"state_snapshot","roles":[],"inserted":false,"precision":{"weight":null,"activation":"f32","output":null},"steps":["state f32"],"group":416,"kernel":"causal_conv1d::causal_conv1d_update_l2norm_f32","family":"causal_conv1d_l2norm","point":null,"compute":"cuda_core","numerics":"reference","where":"device","state":"conv:snapshot"},{"id":"l29.gdn.l2","order":688,"layer":29,"block":"l29.gdn","block_type":"gdn","kind":"linear_attention","op":"l2_norm","roles":[],"inserted":false,"precision":{"weight":null,"activation":"f32","output":"f32"},"steps":["compute f32"],"group":416,"kernel":"causal_conv1d::causal_conv1d_update_l2norm_f32","family":"causal_conv1d_l2norm","point":null,"compute":"cuda_core","numerics":"reference","where":"device","state":null},{"id":"l29.gdn.recur","order":689,"layer":29,"block":"l29.gdn","block_type":"gdn","kind":"linear_attention","op":"gdn_recurrence","roles":[],"inserted":false,"precision":{"weight":null,"activation":"f32","output":"f32"},"steps":["state f32","compute f32"],"group":417,"kernel":"gated_delta_rule::gated_delta_rule_decode_f32","family":"gdn_recurrence","point":{"decay":"head","gate_act":"silu"},"compute":"cuda_core","numerics":"reference","where":"device","state":"h:update"},{"id":"l29.gdn.out_norm","order":690,"layer":29,"block":"l29.gdn","block_type":"gdn","kind":"linear_attention","op":"gated_rms_norm","roles":[],"inserted":false,"precision":{"weight":null,"activation":"f32","output":"bf16"},"steps":["compute f32"],"group":418,"kernel":"norm::gated_rms_norm_f32_input","family":"gated_rms_norm","point":null,"compute":"memory","numerics":"reference","where":"device","state":null},{"id":"l29.gdn.out","order":691,"layer":29,"block":"l29.gdn","block_type":"gdn","kind":"linear_attention","op":"linear:gdn_out","roles":["gdn_out"],"inserted":false,"precision":{"weight":"fp8/block128x128","activation":"bf16","output":"bf16"},"steps":["act bf16","weight fp8/block128x128->bf16","mma bf16*bf16","accumulate f32","scale f32"],"group":419,"kernel":"w8a16_gemm_pipelined_m32::w8a16_gemm_pipelined_m32","family":"w8a16_gemm","point":null,"compute":"tensor_core:mma.sync.m16n8k16.bf16","numerics":"reference","where":"device","state":null},{"id":"l29.gdn.add","order":692,"layer":29,"block":"l29.gdn","block_type":"gdn","kind":"linear_attention","op":"residual_add","roles":[],"inserted":false,"precision":{"weight":null,"activation":"bf16","output":"bf16"},"steps":["compute f32"],"group":420,"kernel":"norm::residual_add_rms_norm","family":"rms_norm","point":null,"compute":"memory","numerics":"reference","where":"device","state":null},{"id":"l29.moe_ffn.post_norm","order":693,"layer":29,"block":"l29.moe_ffn","block_type":"moe_ffn","kind":"linear_attention","op":"rms_norm","roles":[],"inserted":false,"precision":{"weight":null,"activation":"f32","output":"bf16"},"steps":["compute f32"],"group":420,"kernel":"norm::residual_add_rms_norm","family":"rms_norm","point":null,"compute":"memory","numerics":"reference","where":"device","state":null},{"id":"l29.moe_ffn.router","order":694,"layer":29,"block":"l29.moe_ffn","block_type":"moe_ffn","kind":"linear_attention","op":"router","roles":[],"inserted":false,"precision":{"weight":"bf16","activation":"bf16","output":"bf16"},"steps":["act bf16","weight bf16->bf16","mma bf16*bf16","accumulate f32","scale none"],"group":421,"kernel":"gemv::dense_gemv_bf16","family":"dense_bf16","point":null,"compute":"cuda_core","numerics":"reference","where":"device","state":null},{"id":"l29.moe_ffn.top_k","order":695,"layer":29,"block":"l29.moe_ffn","block_type":"moe_ffn","kind":"linear_attention","op":"top_k","roles":[],"inserted":false,"precision":{"weight":null,"activation":"bf16","output":"f32,i32"},"steps":["score f32"],"group":422,"kernel":"moe_topk::moe_topk_softmax","family":"moe_topk","point":null,"compute":"memory","numerics":"reference","where":"device","state":null},{"id":"l29.moe_ffn.experts_gate_up","order":696,"layer":29,"block":"l29.moe_ffn","block_type":"moe_ffn","kind":"linear_attention","op":"expert_gate_up","roles":[],"inserted":false,"precision":{"weight":"fp8/block128x128","activation":"bf16","output":"bf16"},"steps":["gather bf16","act bf16","weight fp8/block128x128->bf16","mma bf16*bf16","accumulate f32","scale f32"],"group":423,"kernel":"moe_shared_expert_fused_fp8::moe_expert_gate_up_shared_fp8","family":"moe_fp8_1row","point":null,"compute":"cuda_core","numerics":"reference","where":"device","state":null},{"id":"l29.moe_ffn.shared_gate_up","order":697,"layer":29,"block":"l29.moe_ffn","block_type":"moe_ffn","kind":"linear_attention","op":"linear:shared_gate_up","roles":["shared_gate_up"],"inserted":false,"precision":{"weight":"fp8/block128x128","activation":"bf16","output":"bf16"},"steps":["act bf16","weight fp8/block128x128->bf16","mma bf16*bf16","accumulate f32","scale f32"],"group":423,"kernel":"moe_shared_expert_fused_fp8::moe_expert_gate_up_shared_fp8","family":"moe_fp8_1row","point":null,"compute":"cuda_core","numerics":"reference","where":"device","state":null},{"id":"l29.moe_ffn.experts_act","order":698,"layer":29,"block":"l29.moe_ffn","block_type":"moe_ffn","kind":"linear_attention","op":"silu_mul","roles":[],"inserted":false,"precision":{"weight":null,"activation":"bf16","output":"f32"},"steps":["compute f32"],"group":424,"kernel":"moe_shared_expert_fused_fp8::moe_expert_silu_down_shared_fp8","family":"moe_fp8_1row","point":null,"compute":"cuda_core","numerics":"reference","where":"device","state":null},{"id":"l29.moe_ffn.experts_down","order":699,"layer":29,"block":"l29.moe_ffn","block_type":"moe_ffn","kind":"linear_attention","op":"expert_down","roles":[],"inserted":false,"precision":{"weight":"fp8/block128x128","activation":"f32","output":"bf16"},"steps":["act f32","weight fp8/block128x128->f32","mma f32*f32","accumulate f32","scale f32"],"group":424,"kernel":"moe_shared_expert_fused_fp8::moe_expert_silu_down_shared_fp8","family":"moe_fp8_1row","point":null,"compute":"cuda_core","numerics":"reference","where":"device","state":null},{"id":"l29.moe_ffn.shared_act","order":700,"layer":29,"block":"l29.moe_ffn","block_type":"moe_ffn","kind":"linear_attention","op":"silu_mul","roles":[],"inserted":false,"precision":{"weight":null,"activation":"bf16","output":"f32"},"steps":["compute f32"],"group":424,"kernel":"moe_shared_expert_fused_fp8::moe_expert_silu_down_shared_fp8","family":"moe_fp8_1row","point":null,"compute":"cuda_core","numerics":"reference","where":"device","state":null},{"id":"l29.moe_ffn.shared_down","order":701,"layer":29,"block":"l29.moe_ffn","block_type":"moe_ffn","kind":"linear_attention","op":"linear:shared_down","roles":["shared_down"],"inserted":false,"precision":{"weight":"fp8/block128x128","activation":"f32","output":"bf16"},"steps":["act f32","weight fp8/block128x128->f32","mma f32*f32","accumulate f32","scale f32"],"group":424,"kernel":"moe_shared_expert_fused_fp8::moe_expert_silu_down_shared_fp8","family":"moe_fp8_1row","point":null,"compute":"cuda_core","numerics":"reference","where":"device","state":null},{"id":"l29.moe_ffn.shared_gate","order":702,"layer":29,"block":"l29.moe_ffn","block_type":"moe_ffn","kind":"linear_attention","op":"linear:shared_gate","roles":["shared_gate"],"inserted":false,"precision":{"weight":"bf16","activation":"bf16","output":"f32"},"steps":["act bf16","weight bf16->bf16","mma bf16*bf16","accumulate f32","scale none"],"group":425,"kernel":"moe_expert_gemv::moe_weighted_sum_blend","family":"moe_blend","point":{"shared_gate":"sigmoid"},"compute":"cuda_core","numerics":"reference","where":"device","state":null},{"id":"l29.moe_ffn.blend","order":703,"layer":29,"block":"l29.moe_ffn","block_type":"moe_ffn","kind":"linear_attention","op":"blend","roles":[],"inserted":false,"precision":{"weight":null,"activation":"bf16","output":"bf16"},"steps":["scatter f32","combine f32"],"group":425,"kernel":"moe_expert_gemv::moe_weighted_sum_blend","family":"moe_blend","point":{"shared_gate":"sigmoid"},"compute":"cuda_core","numerics":"reference","where":"device","state":null},{"id":"l29.moe_ffn.add","order":704,"layer":29,"block":"l29.moe_ffn","block_type":"moe_ffn","kind":"linear_attention","op":"residual_add","roles":[],"inserted":false,"precision":{"weight":null,"activation":"bf16","output":"bf16"},"steps":["compute f32"],"group":426,"kernel":"residual_add_rms_norm_exact::residual_add_rms_norm_exact","family":"rms_norm","point":null,"compute":"memory","numerics":"bit_identical(residual_add_rms_norm_exact_microtest)","where":"device","state":null},{"id":"l30.gdn.input_norm","order":705,"layer":30,"block":"l30.gdn","block_type":"gdn","kind":"linear_attention","op":"rms_norm","roles":[],"inserted":false,"precision":{"weight":null,"activation":"bf16","output":"bf16"},"steps":["compute f32"],"group":426,"kernel":"residual_add_rms_norm_exact::residual_add_rms_norm_exact","family":"rms_norm","point":null,"compute":"memory","numerics":"bit_identical(residual_add_rms_norm_exact_microtest)","where":"device","state":null},{"id":"l30.gdn.qkvz","order":706,"layer":30,"block":"l30.gdn","block_type":"gdn","kind":"linear_attention","op":"linear:qkvz","roles":["qkvz"],"inserted":false,"precision":{"weight":"fp8/block128x128","activation":"bf16","output":"bf16"},"steps":["act bf16","weight fp8/block128x128->bf16","mma bf16*bf16","accumulate f32","scale f32"],"group":427,"kernel":"w8a16_gemm_pipelined_m32::w8a16_gemm_pipelined_m32","family":"w8a16_gemm","point":null,"compute":"tensor_core:mma.sync.m16n8k16.bf16","numerics":"reference","where":"device","state":null},{"id":"l30.gdn.ba","order":707,"layer":30,"block":"l30.gdn","block_type":"gdn","kind":"linear_attention","op":"linear:ba","roles":["ba"],"inserted":false,"precision":{"weight":"bf16","activation":"bf16","output":"f32"},"steps":["act bf16","weight bf16->bf16","mma bf16*bf16","accumulate f32","scale none"],"group":428,"kernel":"ssm_preprocess::dense_gemv_ba_gates","family":"gdn_ba_gates_gemv","point":{"gate":"softplus","decay":"head"},"compute":"cuda_core","numerics":"reference","where":"device","state":null},{"id":"l30.gdn.gates","order":708,"layer":30,"block":"l30.gdn","block_type":"gdn","kind":"linear_attention","op":"gdn_gates","roles":[],"inserted":false,"precision":{"weight":null,"activation":"f32","output":"f32,f32"},"steps":["compute f32"],"group":428,"kernel":"ssm_preprocess::dense_gemv_ba_gates","family":"gdn_ba_gates_gemv","point":{"gate":"softplus","decay":"head"},"compute":"cuda_core","numerics":"reference","where":"device","state":null},{"id":"l30.gdn.conv","order":709,"layer":30,"block":"l30.gdn","block_type":"gdn","kind":"linear_attention","op":"conv1d_update","roles":[],"inserted":false,"precision":{"weight":null,"activation":"bf16","output":"f32"},"steps":["state f32","compute f32"],"group":429,"kernel":"causal_conv1d::causal_conv1d_update_l2norm_f32","family":"causal_conv1d_l2norm","point":null,"compute":"cuda_core","numerics":"reference","where":"device","state":"conv:update"},{"id":"l30.gdn.conv_ckpt","order":710,"layer":30,"block":"l30.gdn","block_type":"gdn","kind":"linear_attention","op":"state_snapshot","roles":[],"inserted":false,"precision":{"weight":null,"activation":"f32","output":null},"steps":["state f32"],"group":429,"kernel":"causal_conv1d::causal_conv1d_update_l2norm_f32","family":"causal_conv1d_l2norm","point":null,"compute":"cuda_core","numerics":"reference","where":"device","state":"conv:snapshot"},{"id":"l30.gdn.l2","order":711,"layer":30,"block":"l30.gdn","block_type":"gdn","kind":"linear_attention","op":"l2_norm","roles":[],"inserted":false,"precision":{"weight":null,"activation":"f32","output":"f32"},"steps":["compute f32"],"group":429,"kernel":"causal_conv1d::causal_conv1d_update_l2norm_f32","family":"causal_conv1d_l2norm","point":null,"compute":"cuda_core","numerics":"reference","where":"device","state":null},{"id":"l30.gdn.recur","order":712,"layer":30,"block":"l30.gdn","block_type":"gdn","kind":"linear_attention","op":"gdn_recurrence","roles":[],"inserted":false,"precision":{"weight":null,"activation":"f32","output":"f32"},"steps":["state f32","compute f32"],"group":430,"kernel":"gated_delta_rule::gated_delta_rule_decode_f32","family":"gdn_recurrence","point":{"decay":"head","gate_act":"silu"},"compute":"cuda_core","numerics":"reference","where":"device","state":"h:update"},{"id":"l30.gdn.out_norm","order":713,"layer":30,"block":"l30.gdn","block_type":"gdn","kind":"linear_attention","op":"gated_rms_norm","roles":[],"inserted":false,"precision":{"weight":null,"activation":"f32","output":"bf16"},"steps":["compute f32"],"group":431,"kernel":"norm::gated_rms_norm_f32_input","family":"gated_rms_norm","point":null,"compute":"memory","numerics":"reference","where":"device","state":null},{"id":"l30.gdn.out","order":714,"layer":30,"block":"l30.gdn","block_type":"gdn","kind":"linear_attention","op":"linear:gdn_out","roles":["gdn_out"],"inserted":false,"precision":{"weight":"fp8/block128x128","activation":"bf16","output":"bf16"},"steps":["act bf16","weight fp8/block128x128->bf16","mma bf16*bf16","accumulate f32","scale f32"],"group":432,"kernel":"w8a16_gemm_pipelined_m32::w8a16_gemm_pipelined_m32","family":"w8a16_gemm","point":null,"compute":"tensor_core:mma.sync.m16n8k16.bf16","numerics":"reference","where":"device","state":null},{"id":"l30.gdn.add","order":715,"layer":30,"block":"l30.gdn","block_type":"gdn","kind":"linear_attention","op":"residual_add","roles":[],"inserted":false,"precision":{"weight":null,"activation":"bf16","output":"bf16"},"steps":["compute f32"],"group":433,"kernel":"norm::residual_add_rms_norm","family":"rms_norm","point":null,"compute":"memory","numerics":"reference","where":"device","state":null},{"id":"l30.moe_ffn.post_norm","order":716,"layer":30,"block":"l30.moe_ffn","block_type":"moe_ffn","kind":"linear_attention","op":"rms_norm","roles":[],"inserted":false,"precision":{"weight":null,"activation":"f32","output":"bf16"},"steps":["compute f32"],"group":433,"kernel":"norm::residual_add_rms_norm","family":"rms_norm","point":null,"compute":"memory","numerics":"reference","where":"device","state":null},{"id":"l30.moe_ffn.router","order":717,"layer":30,"block":"l30.moe_ffn","block_type":"moe_ffn","kind":"linear_attention","op":"router","roles":[],"inserted":false,"precision":{"weight":"bf16","activation":"bf16","output":"bf16"},"steps":["act bf16","weight bf16->bf16","mma bf16*bf16","accumulate f32","scale none"],"group":434,"kernel":"gemv::dense_gemv_bf16","family":"dense_bf16","point":null,"compute":"cuda_core","numerics":"reference","where":"device","state":null},{"id":"l30.moe_ffn.top_k","order":718,"layer":30,"block":"l30.moe_ffn","block_type":"moe_ffn","kind":"linear_attention","op":"top_k","roles":[],"inserted":false,"precision":{"weight":null,"activation":"bf16","output":"f32,i32"},"steps":["score f32"],"group":435,"kernel":"moe_topk::moe_topk_softmax","family":"moe_topk","point":null,"compute":"memory","numerics":"reference","where":"device","state":null},{"id":"l30.moe_ffn.experts_gate_up","order":719,"layer":30,"block":"l30.moe_ffn","block_type":"moe_ffn","kind":"linear_attention","op":"expert_gate_up","roles":[],"inserted":false,"precision":{"weight":"fp8/block128x128","activation":"bf16","output":"bf16"},"steps":["gather bf16","act bf16","weight fp8/block128x128->bf16","mma bf16*bf16","accumulate f32","scale f32"],"group":436,"kernel":"moe_shared_expert_fused_fp8::moe_expert_gate_up_shared_fp8","family":"moe_fp8_1row","point":null,"compute":"cuda_core","numerics":"reference","where":"device","state":null},{"id":"l30.moe_ffn.shared_gate_up","order":720,"layer":30,"block":"l30.moe_ffn","block_type":"moe_ffn","kind":"linear_attention","op":"linear:shared_gate_up","roles":["shared_gate_up"],"inserted":false,"precision":{"weight":"fp8/block128x128","activation":"bf16","output":"bf16"},"steps":["act bf16","weight fp8/block128x128->bf16","mma bf16*bf16","accumulate f32","scale f32"],"group":436,"kernel":"moe_shared_expert_fused_fp8::moe_expert_gate_up_shared_fp8","family":"moe_fp8_1row","point":null,"compute":"cuda_core","numerics":"reference","where":"device","state":null},{"id":"l30.moe_ffn.experts_act","order":721,"layer":30,"block":"l30.moe_ffn","block_type":"moe_ffn","kind":"linear_attention","op":"silu_mul","roles":[],"inserted":false,"precision":{"weight":null,"activation":"bf16","output":"f32"},"steps":["compute f32"],"group":437,"kernel":"moe_shared_expert_fused_fp8::moe_expert_silu_down_shared_fp8","family":"moe_fp8_1row","point":null,"compute":"cuda_core","numerics":"reference","where":"device","state":null},{"id":"l30.moe_ffn.experts_down","order":722,"layer":30,"block":"l30.moe_ffn","block_type":"moe_ffn","kind":"linear_attention","op":"expert_down","roles":[],"inserted":false,"precision":{"weight":"fp8/block128x128","activation":"f32","output":"bf16"},"steps":["act f32","weight fp8/block128x128->f32","mma f32*f32","accumulate f32","scale f32"],"group":437,"kernel":"moe_shared_expert_fused_fp8::moe_expert_silu_down_shared_fp8","family":"moe_fp8_1row","point":null,"compute":"cuda_core","numerics":"reference","where":"device","state":null},{"id":"l30.moe_ffn.shared_act","order":723,"layer":30,"block":"l30.moe_ffn","block_type":"moe_ffn","kind":"linear_attention","op":"silu_mul","roles":[],"inserted":false,"precision":{"weight":null,"activation":"bf16","output":"f32"},"steps":["compute f32"],"group":437,"kernel":"moe_shared_expert_fused_fp8::moe_expert_silu_down_shared_fp8","family":"moe_fp8_1row","point":null,"compute":"cuda_core","numerics":"reference","where":"device","state":null},{"id":"l30.moe_ffn.shared_down","order":724,"layer":30,"block":"l30.moe_ffn","block_type":"moe_ffn","kind":"linear_attention","op":"linear:shared_down","roles":["shared_down"],"inserted":false,"precision":{"weight":"fp8/block128x128","activation":"f32","output":"bf16"},"steps":["act f32","weight fp8/block128x128->f32","mma f32*f32","accumulate f32","scale f32"],"group":437,"kernel":"moe_shared_expert_fused_fp8::moe_expert_silu_down_shared_fp8","family":"moe_fp8_1row","point":null,"compute":"cuda_core","numerics":"reference","where":"device","state":null},{"id":"l30.moe_ffn.shared_gate","order":725,"layer":30,"block":"l30.moe_ffn","block_type":"moe_ffn","kind":"linear_attention","op":"linear:shared_gate","roles":["shared_gate"],"inserted":false,"precision":{"weight":"bf16","activation":"bf16","output":"f32"},"steps":["act bf16","weight bf16->bf16","mma bf16*bf16","accumulate f32","scale none"],"group":438,"kernel":"moe_expert_gemv::moe_weighted_sum_blend","family":"moe_blend","point":{"shared_gate":"sigmoid"},"compute":"cuda_core","numerics":"reference","where":"device","state":null},{"id":"l30.moe_ffn.blend","order":726,"layer":30,"block":"l30.moe_ffn","block_type":"moe_ffn","kind":"linear_attention","op":"blend","roles":[],"inserted":false,"precision":{"weight":null,"activation":"bf16","output":"bf16"},"steps":["scatter f32","combine f32"],"group":438,"kernel":"moe_expert_gemv::moe_weighted_sum_blend","family":"moe_blend","point":{"shared_gate":"sigmoid"},"compute":"cuda_core","numerics":"reference","where":"device","state":null},{"id":"l30.moe_ffn.add","order":727,"layer":30,"block":"l30.moe_ffn","block_type":"moe_ffn","kind":"linear_attention","op":"residual_add","roles":[],"inserted":false,"precision":{"weight":null,"activation":"bf16","output":"bf16"},"steps":["compute f32"],"group":439,"kernel":"residual_add_rms_norm_exact::residual_add_rms_norm_exact","family":"rms_norm","point":null,"compute":"memory","numerics":"bit_identical(residual_add_rms_norm_exact_microtest)","where":"device","state":null},{"id":"l31.attn.input_norm","order":728,"layer":31,"block":"l31.attn","block_type":"attn","kind":"full_attention","op":"rms_norm","roles":[],"inserted":false,"precision":{"weight":null,"activation":"bf16","output":"bf16"},"steps":["compute f32"],"group":439,"kernel":"residual_add_rms_norm_exact::residual_add_rms_norm_exact","family":"rms_norm","point":null,"compute":"memory","numerics":"bit_identical(residual_add_rms_norm_exact_microtest)","where":"device","state":null},{"id":"l31.attn.q","order":729,"layer":31,"block":"l31.attn","block_type":"attn","kind":"full_attention","op":"linear:q","roles":["q"],"inserted":false,"precision":{"weight":"fp8/block128x128","activation":"bf16","output":"bf16"},"steps":["act bf16","weight fp8/block128x128->bf16","mma bf16*bf16","accumulate f32","scale f32"],"group":440,"kernel":"w8a16_gemm_pipelined_m32::w8a16_gemm_pipelined_m32","family":"w8a16_gemm","point":null,"compute":"tensor_core:mma.sync.m16n8k16.bf16","numerics":"reference","where":"device","state":null},{"id":"l31.attn.q_split","order":730,"layer":31,"block":"l31.attn","block_type":"attn","kind":"full_attention","op":"split","roles":[],"inserted":false,"precision":{"weight":null,"activation":"bf16","output":"bf16,bf16"},"steps":["move bf16"],"group":441,"kernel":"ssm_preprocess::deinterleave_qg","family":"deinterleave_qg","point":{"layout":"interleaved"},"compute":"memory","numerics":"reference","where":"device","state":null},{"id":"l31.attn.k","order":731,"layer":31,"block":"l31.attn","block_type":"attn","kind":"full_attention","op":"linear:k","roles":["k"],"inserted":false,"precision":{"weight":"fp8/block128x128","activation":"bf16","output":"bf16"},"steps":["act bf16","weight fp8/block128x128->bf16","mma bf16*bf16","accumulate f32","scale f32"],"group":442,"kernel":"w8a16_gemm_pipelined_m32::w8a16_gemm_pipelined_m32","family":"w8a16_gemm","point":null,"compute":"tensor_core:mma.sync.m16n8k16.bf16","numerics":"reference","where":"device","state":null},{"id":"l31.attn.v","order":732,"layer":31,"block":"l31.attn","block_type":"attn","kind":"full_attention","op":"linear:v","roles":["v"],"inserted":false,"precision":{"weight":"fp8/block128x128","activation":"bf16","output":"bf16"},"steps":["act bf16","weight fp8/block128x128->bf16","mma bf16*bf16","accumulate f32","scale f32"],"group":443,"kernel":"w8a16_gemm_pipelined_m32::w8a16_gemm_pipelined_m32","family":"w8a16_gemm","point":null,"compute":"tensor_core:mma.sync.m16n8k16.bf16","numerics":"reference","where":"device","state":null},{"id":"l31.attn.q_norm","order":733,"layer":31,"block":"l31.attn","block_type":"attn","kind":"full_attention","op":"qk_norm","roles":[],"inserted":false,"precision":{"weight":null,"activation":"bf16","output":"bf16"},"steps":["compute f32"],"group":444,"kernel":"norm::rms_norm","family":"rms_norm","point":null,"compute":"memory","numerics":"reference","where":"device","state":null},{"id":"l31.attn.k_norm","order":734,"layer":31,"block":"l31.attn","block_type":"attn","kind":"full_attention","op":"qk_norm","roles":[],"inserted":false,"precision":{"weight":null,"activation":"bf16","output":"bf16"},"steps":["compute f32"],"group":445,"kernel":"norm::rms_norm","family":"rms_norm","point":null,"compute":"memory","numerics":"reference","where":"device","state":null},{"id":"l31.attn.rope","order":735,"layer":31,"block":"l31.attn","block_type":"attn","kind":"full_attention","op":"rope","roles":[],"inserted":false,"precision":{"weight":null,"activation":"bf16","output":"bf16,bf16"},"steps":["compute f32"],"group":446,"kernel":"rope_mrope_interleaved::rope_forward_mrope_interleaved","family":"rope","point":null,"compute":"memory","numerics":"reference","where":"device","state":null},{"id":"l31.attn.kv_write","order":736,"layer":31,"block":"l31.attn","block_type":"attn","kind":"full_attention","op":"kv_write","roles":[],"inserted":false,"precision":{"weight":null,"activation":"bf16","output":null},"steps":["cache bf16"],"group":447,"kernel":"reshape_and_cache::reshape_and_cache_flash","family":"kv_write","point":null,"compute":"memory","numerics":"reference","where":"device","state":"k:write,v:write"},{"id":"l31.attn.attend","order":737,"layer":31,"block":"l31.attn","block_type":"attn","kind":"full_attention","op":"paged_attention","roles":[],"inserted":false,"precision":{"weight":null,"activation":"bf16","output":"bf16"},"steps":["cache bf16","scores f32","softmax f32","accumulate f32"],"group":448,"kernel":"paged_decode::paged_decode_attn","family":"paged_decode_attn","point":null,"compute":"cuda_core","numerics":"reference","where":"device","state":"k:read,v:read"},{"id":"l31.attn.gate_mul","order":738,"layer":31,"block":"l31.attn","block_type":"attn","kind":"full_attention","op":"sigmoid_gate_mul","roles":[],"inserted":false,"precision":{"weight":null,"activation":"bf16","output":"bf16"},"steps":["compute f32"],"group":449,"kernel":"residual_add::sigmoid_gate_mul","family":"sigmoid_gate_mul","point":null,"compute":"memory","numerics":"reference","where":"device","state":null},{"id":"l31.attn.o","order":739,"layer":31,"block":"l31.attn","block_type":"attn","kind":"full_attention","op":"linear:o","roles":["o"],"inserted":false,"precision":{"weight":"fp8/block128x128","activation":"bf16","output":"bf16"},"steps":["act bf16","weight fp8/block128x128->bf16","mma bf16*bf16","accumulate f32","scale f32"],"group":450,"kernel":"w8a16_gemm_pipelined_m32::w8a16_gemm_pipelined_m32","family":"w8a16_gemm","point":null,"compute":"tensor_core:mma.sync.m16n8k16.bf16","numerics":"reference","where":"device","state":null},{"id":"l31.attn.add","order":740,"layer":31,"block":"l31.attn","block_type":"attn","kind":"full_attention","op":"residual_add","roles":[],"inserted":false,"precision":{"weight":null,"activation":"bf16","output":"bf16"},"steps":["compute f32"],"group":451,"kernel":"norm::residual_add_rms_norm","family":"rms_norm","point":null,"compute":"memory","numerics":"reference","where":"device","state":null},{"id":"l31.moe_ffn.post_norm","order":741,"layer":31,"block":"l31.moe_ffn","block_type":"moe_ffn","kind":"full_attention","op":"rms_norm","roles":[],"inserted":false,"precision":{"weight":null,"activation":"f32","output":"bf16"},"steps":["compute f32"],"group":451,"kernel":"norm::residual_add_rms_norm","family":"rms_norm","point":null,"compute":"memory","numerics":"reference","where":"device","state":null},{"id":"l31.moe_ffn.router","order":742,"layer":31,"block":"l31.moe_ffn","block_type":"moe_ffn","kind":"full_attention","op":"router","roles":[],"inserted":false,"precision":{"weight":"bf16","activation":"bf16","output":"bf16"},"steps":["act bf16","weight bf16->bf16","mma bf16*bf16","accumulate f32","scale none"],"group":452,"kernel":"gemv::dense_gemv_bf16","family":"dense_bf16","point":null,"compute":"cuda_core","numerics":"reference","where":"device","state":null},{"id":"l31.moe_ffn.top_k","order":743,"layer":31,"block":"l31.moe_ffn","block_type":"moe_ffn","kind":"full_attention","op":"top_k","roles":[],"inserted":false,"precision":{"weight":null,"activation":"bf16","output":"f32,i32"},"steps":["score f32"],"group":453,"kernel":"moe_topk::moe_topk_softmax","family":"moe_topk","point":null,"compute":"memory","numerics":"reference","where":"device","state":null},{"id":"l31.moe_ffn.experts_gate_up","order":744,"layer":31,"block":"l31.moe_ffn","block_type":"moe_ffn","kind":"full_attention","op":"expert_gate_up","roles":[],"inserted":false,"precision":{"weight":"fp8/block128x128","activation":"bf16","output":"bf16"},"steps":["gather bf16","act bf16","weight fp8/block128x128->bf16","mma bf16*bf16","accumulate f32","scale f32"],"group":454,"kernel":"moe_shared_expert_fused_fp8::moe_expert_gate_up_shared_fp8","family":"moe_fp8_1row","point":null,"compute":"cuda_core","numerics":"reference","where":"device","state":null},{"id":"l31.moe_ffn.shared_gate_up","order":745,"layer":31,"block":"l31.moe_ffn","block_type":"moe_ffn","kind":"full_attention","op":"linear:shared_gate_up","roles":["shared_gate_up"],"inserted":false,"precision":{"weight":"fp8/block128x128","activation":"bf16","output":"bf16"},"steps":["act bf16","weight fp8/block128x128->bf16","mma bf16*bf16","accumulate f32","scale f32"],"group":454,"kernel":"moe_shared_expert_fused_fp8::moe_expert_gate_up_shared_fp8","family":"moe_fp8_1row","point":null,"compute":"cuda_core","numerics":"reference","where":"device","state":null},{"id":"l31.moe_ffn.experts_act","order":746,"layer":31,"block":"l31.moe_ffn","block_type":"moe_ffn","kind":"full_attention","op":"silu_mul","roles":[],"inserted":false,"precision":{"weight":null,"activation":"bf16","output":"f32"},"steps":["compute f32"],"group":455,"kernel":"moe_shared_expert_fused_fp8::moe_expert_silu_down_shared_fp8","family":"moe_fp8_1row","point":null,"compute":"cuda_core","numerics":"reference","where":"device","state":null},{"id":"l31.moe_ffn.experts_down","order":747,"layer":31,"block":"l31.moe_ffn","block_type":"moe_ffn","kind":"full_attention","op":"expert_down","roles":[],"inserted":false,"precision":{"weight":"fp8/block128x128","activation":"f32","output":"bf16"},"steps":["act f32","weight fp8/block128x128->f32","mma f32*f32","accumulate f32","scale f32"],"group":455,"kernel":"moe_shared_expert_fused_fp8::moe_expert_silu_down_shared_fp8","family":"moe_fp8_1row","point":null,"compute":"cuda_core","numerics":"reference","where":"device","state":null},{"id":"l31.moe_ffn.shared_act","order":748,"layer":31,"block":"l31.moe_ffn","block_type":"moe_ffn","kind":"full_attention","op":"silu_mul","roles":[],"inserted":false,"precision":{"weight":null,"activation":"bf16","output":"f32"},"steps":["compute f32"],"group":455,"kernel":"moe_shared_expert_fused_fp8::moe_expert_silu_down_shared_fp8","family":"moe_fp8_1row","point":null,"compute":"cuda_core","numerics":"reference","where":"device","state":null},{"id":"l31.moe_ffn.shared_down","order":749,"layer":31,"block":"l31.moe_ffn","block_type":"moe_ffn","kind":"full_attention","op":"linear:shared_down","roles":["shared_down"],"inserted":false,"precision":{"weight":"fp8/block128x128","activation":"f32","output":"bf16"},"steps":["act f32","weight fp8/block128x128->f32","mma f32*f32","accumulate f32","scale f32"],"group":455,"kernel":"moe_shared_expert_fused_fp8::moe_expert_silu_down_shared_fp8","family":"moe_fp8_1row","point":null,"compute":"cuda_core","numerics":"reference","where":"device","state":null},{"id":"l31.moe_ffn.shared_gate","order":750,"layer":31,"block":"l31.moe_ffn","block_type":"moe_ffn","kind":"full_attention","op":"linear:shared_gate","roles":["shared_gate"],"inserted":false,"precision":{"weight":"bf16","activation":"bf16","output":"f32"},"steps":["act bf16","weight bf16->bf16","mma bf16*bf16","accumulate f32","scale none"],"group":456,"kernel":"moe_expert_gemv::moe_weighted_sum_blend","family":"moe_blend","point":{"shared_gate":"sigmoid"},"compute":"cuda_core","numerics":"reference","where":"device","state":null},{"id":"l31.moe_ffn.blend","order":751,"layer":31,"block":"l31.moe_ffn","block_type":"moe_ffn","kind":"full_attention","op":"blend","roles":[],"inserted":false,"precision":{"weight":null,"activation":"bf16","output":"bf16"},"steps":["scatter f32","combine f32"],"group":456,"kernel":"moe_expert_gemv::moe_weighted_sum_blend","family":"moe_blend","point":{"shared_gate":"sigmoid"},"compute":"cuda_core","numerics":"reference","where":"device","state":null},{"id":"l31.moe_ffn.add","order":752,"layer":31,"block":"l31.moe_ffn","block_type":"moe_ffn","kind":"full_attention","op":"residual_add","roles":[],"inserted":false,"precision":{"weight":null,"activation":"bf16","output":"bf16"},"steps":["compute f32"],"group":457,"kernel":"residual_add_rms_norm_exact::residual_add_rms_norm_exact","family":"rms_norm","point":null,"compute":"memory","numerics":"bit_identical(residual_add_rms_norm_exact_microtest)","where":"device","state":null},{"id":"l32.gdn.input_norm","order":753,"layer":32,"block":"l32.gdn","block_type":"gdn","kind":"linear_attention","op":"rms_norm","roles":[],"inserted":false,"precision":{"weight":null,"activation":"bf16","output":"bf16"},"steps":["compute f32"],"group":457,"kernel":"residual_add_rms_norm_exact::residual_add_rms_norm_exact","family":"rms_norm","point":null,"compute":"memory","numerics":"bit_identical(residual_add_rms_norm_exact_microtest)","where":"device","state":null},{"id":"l32.gdn.qkvz","order":754,"layer":32,"block":"l32.gdn","block_type":"gdn","kind":"linear_attention","op":"linear:qkvz","roles":["qkvz"],"inserted":false,"precision":{"weight":"fp8/block128x128","activation":"bf16","output":"bf16"},"steps":["act bf16","weight fp8/block128x128->bf16","mma bf16*bf16","accumulate f32","scale f32"],"group":458,"kernel":"w8a16_gemm_pipelined_m32::w8a16_gemm_pipelined_m32","family":"w8a16_gemm","point":null,"compute":"tensor_core:mma.sync.m16n8k16.bf16","numerics":"reference","where":"device","state":null},{"id":"l32.gdn.ba","order":755,"layer":32,"block":"l32.gdn","block_type":"gdn","kind":"linear_attention","op":"linear:ba","roles":["ba"],"inserted":false,"precision":{"weight":"bf16","activation":"bf16","output":"f32"},"steps":["act bf16","weight bf16->bf16","mma bf16*bf16","accumulate f32","scale none"],"group":459,"kernel":"ssm_preprocess::dense_gemv_ba_gates","family":"gdn_ba_gates_gemv","point":{"gate":"softplus","decay":"head"},"compute":"cuda_core","numerics":"reference","where":"device","state":null},{"id":"l32.gdn.gates","order":756,"layer":32,"block":"l32.gdn","block_type":"gdn","kind":"linear_attention","op":"gdn_gates","roles":[],"inserted":false,"precision":{"weight":null,"activation":"f32","output":"f32,f32"},"steps":["compute f32"],"group":459,"kernel":"ssm_preprocess::dense_gemv_ba_gates","family":"gdn_ba_gates_gemv","point":{"gate":"softplus","decay":"head"},"compute":"cuda_core","numerics":"reference","where":"device","state":null},{"id":"l32.gdn.conv","order":757,"layer":32,"block":"l32.gdn","block_type":"gdn","kind":"linear_attention","op":"conv1d_update","roles":[],"inserted":false,"precision":{"weight":null,"activation":"bf16","output":"f32"},"steps":["state f32","compute f32"],"group":460,"kernel":"causal_conv1d::causal_conv1d_update_l2norm_f32","family":"causal_conv1d_l2norm","point":null,"compute":"cuda_core","numerics":"reference","where":"device","state":"conv:update"},{"id":"l32.gdn.conv_ckpt","order":758,"layer":32,"block":"l32.gdn","block_type":"gdn","kind":"linear_attention","op":"state_snapshot","roles":[],"inserted":false,"precision":{"weight":null,"activation":"f32","output":null},"steps":["state f32"],"group":460,"kernel":"causal_conv1d::causal_conv1d_update_l2norm_f32","family":"causal_conv1d_l2norm","point":null,"compute":"cuda_core","numerics":"reference","where":"device","state":"conv:snapshot"},{"id":"l32.gdn.l2","order":759,"layer":32,"block":"l32.gdn","block_type":"gdn","kind":"linear_attention","op":"l2_norm","roles":[],"inserted":false,"precision":{"weight":null,"activation":"f32","output":"f32"},"steps":["compute f32"],"group":460,"kernel":"causal_conv1d::causal_conv1d_update_l2norm_f32","family":"causal_conv1d_l2norm","point":null,"compute":"cuda_core","numerics":"reference","where":"device","state":null},{"id":"l32.gdn.recur","order":760,"layer":32,"block":"l32.gdn","block_type":"gdn","kind":"linear_attention","op":"gdn_recurrence","roles":[],"inserted":false,"precision":{"weight":null,"activation":"f32","output":"f32"},"steps":["state f32","compute f32"],"group":461,"kernel":"gated_delta_rule::gated_delta_rule_decode_f32","family":"gdn_recurrence","point":{"decay":"head","gate_act":"silu"},"compute":"cuda_core","numerics":"reference","where":"device","state":"h:update"},{"id":"l32.gdn.out_norm","order":761,"layer":32,"block":"l32.gdn","block_type":"gdn","kind":"linear_attention","op":"gated_rms_norm","roles":[],"inserted":false,"precision":{"weight":null,"activation":"f32","output":"bf16"},"steps":["compute f32"],"group":462,"kernel":"norm::gated_rms_norm_f32_input","family":"gated_rms_norm","point":null,"compute":"memory","numerics":"reference","where":"device","state":null},{"id":"l32.gdn.out","order":762,"layer":32,"block":"l32.gdn","block_type":"gdn","kind":"linear_attention","op":"linear:gdn_out","roles":["gdn_out"],"inserted":false,"precision":{"weight":"fp8/block128x128","activation":"bf16","output":"bf16"},"steps":["act bf16","weight fp8/block128x128->bf16","mma bf16*bf16","accumulate f32","scale f32"],"group":463,"kernel":"w8a16_gemm_pipelined_m32::w8a16_gemm_pipelined_m32","family":"w8a16_gemm","point":null,"compute":"tensor_core:mma.sync.m16n8k16.bf16","numerics":"reference","where":"device","state":null},{"id":"l32.gdn.add","order":763,"layer":32,"block":"l32.gdn","block_type":"gdn","kind":"linear_attention","op":"residual_add","roles":[],"inserted":false,"precision":{"weight":null,"activation":"bf16","output":"bf16"},"steps":["compute f32"],"group":464,"kernel":"norm::residual_add_rms_norm","family":"rms_norm","point":null,"compute":"memory","numerics":"reference","where":"device","state":null},{"id":"l32.moe_ffn.post_norm","order":764,"layer":32,"block":"l32.moe_ffn","block_type":"moe_ffn","kind":"linear_attention","op":"rms_norm","roles":[],"inserted":false,"precision":{"weight":null,"activation":"f32","output":"bf16"},"steps":["compute f32"],"group":464,"kernel":"norm::residual_add_rms_norm","family":"rms_norm","point":null,"compute":"memory","numerics":"reference","where":"device","state":null},{"id":"l32.moe_ffn.router","order":765,"layer":32,"block":"l32.moe_ffn","block_type":"moe_ffn","kind":"linear_attention","op":"router","roles":[],"inserted":false,"precision":{"weight":"bf16","activation":"bf16","output":"bf16"},"steps":["act bf16","weight bf16->bf16","mma bf16*bf16","accumulate f32","scale none"],"group":465,"kernel":"gemv::dense_gemv_bf16","family":"dense_bf16","point":null,"compute":"cuda_core","numerics":"reference","where":"device","state":null},{"id":"l32.moe_ffn.top_k","order":766,"layer":32,"block":"l32.moe_ffn","block_type":"moe_ffn","kind":"linear_attention","op":"top_k","roles":[],"inserted":false,"precision":{"weight":null,"activation":"bf16","output":"f32,i32"},"steps":["score f32"],"group":466,"kernel":"moe_topk::moe_topk_softmax","family":"moe_topk","point":null,"compute":"memory","numerics":"reference","where":"device","state":null},{"id":"l32.moe_ffn.experts_gate_up","order":767,"layer":32,"block":"l32.moe_ffn","block_type":"moe_ffn","kind":"linear_attention","op":"expert_gate_up","roles":[],"inserted":false,"precision":{"weight":"fp8/block128x128","activation":"bf16","output":"bf16"},"steps":["gather bf16","act bf16","weight fp8/block128x128->bf16","mma bf16*bf16","accumulate f32","scale f32"],"group":467,"kernel":"moe_shared_expert_fused_fp8::moe_expert_gate_up_shared_fp8","family":"moe_fp8_1row","point":null,"compute":"cuda_core","numerics":"reference","where":"device","state":null},{"id":"l32.moe_ffn.shared_gate_up","order":768,"layer":32,"block":"l32.moe_ffn","block_type":"moe_ffn","kind":"linear_attention","op":"linear:shared_gate_up","roles":["shared_gate_up"],"inserted":false,"precision":{"weight":"fp8/block128x128","activation":"bf16","output":"bf16"},"steps":["act bf16","weight fp8/block128x128->bf16","mma bf16*bf16","accumulate f32","scale f32"],"group":467,"kernel":"moe_shared_expert_fused_fp8::moe_expert_gate_up_shared_fp8","family":"moe_fp8_1row","point":null,"compute":"cuda_core","numerics":"reference","where":"device","state":null},{"id":"l32.moe_ffn.experts_act","order":769,"layer":32,"block":"l32.moe_ffn","block_type":"moe_ffn","kind":"linear_attention","op":"silu_mul","roles":[],"inserted":false,"precision":{"weight":null,"activation":"bf16","output":"f32"},"steps":["compute f32"],"group":468,"kernel":"moe_shared_expert_fused_fp8::moe_expert_silu_down_shared_fp8","family":"moe_fp8_1row","point":null,"compute":"cuda_core","numerics":"reference","where":"device","state":null},{"id":"l32.moe_ffn.experts_down","order":770,"layer":32,"block":"l32.moe_ffn","block_type":"moe_ffn","kind":"linear_attention","op":"expert_down","roles":[],"inserted":false,"precision":{"weight":"fp8/block128x128","activation":"f32","output":"bf16"},"steps":["act f32","weight fp8/block128x128->f32","mma f32*f32","accumulate f32","scale f32"],"group":468,"kernel":"moe_shared_expert_fused_fp8::moe_expert_silu_down_shared_fp8","family":"moe_fp8_1row","point":null,"compute":"cuda_core","numerics":"reference","where":"device","state":null},{"id":"l32.moe_ffn.shared_act","order":771,"layer":32,"block":"l32.moe_ffn","block_type":"moe_ffn","kind":"linear_attention","op":"silu_mul","roles":[],"inserted":false,"precision":{"weight":null,"activation":"bf16","output":"f32"},"steps":["compute f32"],"group":468,"kernel":"moe_shared_expert_fused_fp8::moe_expert_silu_down_shared_fp8","family":"moe_fp8_1row","point":null,"compute":"cuda_core","numerics":"reference","where":"device","state":null},{"id":"l32.moe_ffn.shared_down","order":772,"layer":32,"block":"l32.moe_ffn","block_type":"moe_ffn","kind":"linear_attention","op":"linear:shared_down","roles":["shared_down"],"inserted":false,"precision":{"weight":"fp8/block128x128","activation":"f32","output":"bf16"},"steps":["act f32","weight fp8/block128x128->f32","mma f32*f32","accumulate f32","scale f32"],"group":468,"kernel":"moe_shared_expert_fused_fp8::moe_expert_silu_down_shared_fp8","family":"moe_fp8_1row","point":null,"compute":"cuda_core","numerics":"reference","where":"device","state":null},{"id":"l32.moe_ffn.shared_gate","order":773,"layer":32,"block":"l32.moe_ffn","block_type":"moe_ffn","kind":"linear_attention","op":"linear:shared_gate","roles":["shared_gate"],"inserted":false,"precision":{"weight":"bf16","activation":"bf16","output":"f32"},"steps":["act bf16","weight bf16->bf16","mma bf16*bf16","accumulate f32","scale none"],"group":469,"kernel":"moe_expert_gemv::moe_weighted_sum_blend","family":"moe_blend","point":{"shared_gate":"sigmoid"},"compute":"cuda_core","numerics":"reference","where":"device","state":null},{"id":"l32.moe_ffn.blend","order":774,"layer":32,"block":"l32.moe_ffn","block_type":"moe_ffn","kind":"linear_attention","op":"blend","roles":[],"inserted":false,"precision":{"weight":null,"activation":"bf16","output":"bf16"},"steps":["scatter f32","combine f32"],"group":469,"kernel":"moe_expert_gemv::moe_weighted_sum_blend","family":"moe_blend","point":{"shared_gate":"sigmoid"},"compute":"cuda_core","numerics":"reference","where":"device","state":null},{"id":"l32.moe_ffn.add","order":775,"layer":32,"block":"l32.moe_ffn","block_type":"moe_ffn","kind":"linear_attention","op":"residual_add","roles":[],"inserted":false,"precision":{"weight":null,"activation":"bf16","output":"bf16"},"steps":["compute f32"],"group":470,"kernel":"residual_add_rms_norm_exact::residual_add_rms_norm_exact","family":"rms_norm","point":null,"compute":"memory","numerics":"bit_identical(residual_add_rms_norm_exact_microtest)","where":"device","state":null},{"id":"l33.gdn.input_norm","order":776,"layer":33,"block":"l33.gdn","block_type":"gdn","kind":"linear_attention","op":"rms_norm","roles":[],"inserted":false,"precision":{"weight":null,"activation":"bf16","output":"bf16"},"steps":["compute f32"],"group":470,"kernel":"residual_add_rms_norm_exact::residual_add_rms_norm_exact","family":"rms_norm","point":null,"compute":"memory","numerics":"bit_identical(residual_add_rms_norm_exact_microtest)","where":"device","state":null},{"id":"l33.gdn.qkvz","order":777,"layer":33,"block":"l33.gdn","block_type":"gdn","kind":"linear_attention","op":"linear:qkvz","roles":["qkvz"],"inserted":false,"precision":{"weight":"fp8/block128x128","activation":"bf16","output":"bf16"},"steps":["act bf16","weight fp8/block128x128->bf16","mma bf16*bf16","accumulate f32","scale f32"],"group":471,"kernel":"w8a16_gemm_pipelined_m32::w8a16_gemm_pipelined_m32","family":"w8a16_gemm","point":null,"compute":"tensor_core:mma.sync.m16n8k16.bf16","numerics":"reference","where":"device","state":null},{"id":"l33.gdn.ba","order":778,"layer":33,"block":"l33.gdn","block_type":"gdn","kind":"linear_attention","op":"linear:ba","roles":["ba"],"inserted":false,"precision":{"weight":"bf16","activation":"bf16","output":"f32"},"steps":["act bf16","weight bf16->bf16","mma bf16*bf16","accumulate f32","scale none"],"group":472,"kernel":"ssm_preprocess::dense_gemv_ba_gates","family":"gdn_ba_gates_gemv","point":{"gate":"softplus","decay":"head"},"compute":"cuda_core","numerics":"reference","where":"device","state":null},{"id":"l33.gdn.gates","order":779,"layer":33,"block":"l33.gdn","block_type":"gdn","kind":"linear_attention","op":"gdn_gates","roles":[],"inserted":false,"precision":{"weight":null,"activation":"f32","output":"f32,f32"},"steps":["compute f32"],"group":472,"kernel":"ssm_preprocess::dense_gemv_ba_gates","family":"gdn_ba_gates_gemv","point":{"gate":"softplus","decay":"head"},"compute":"cuda_core","numerics":"reference","where":"device","state":null},{"id":"l33.gdn.conv","order":780,"layer":33,"block":"l33.gdn","block_type":"gdn","kind":"linear_attention","op":"conv1d_update","roles":[],"inserted":false,"precision":{"weight":null,"activation":"bf16","output":"f32"},"steps":["state f32","compute f32"],"group":473,"kernel":"causal_conv1d::causal_conv1d_update_l2norm_f32","family":"causal_conv1d_l2norm","point":null,"compute":"cuda_core","numerics":"reference","where":"device","state":"conv:update"},{"id":"l33.gdn.conv_ckpt","order":781,"layer":33,"block":"l33.gdn","block_type":"gdn","kind":"linear_attention","op":"state_snapshot","roles":[],"inserted":false,"precision":{"weight":null,"activation":"f32","output":null},"steps":["state f32"],"group":473,"kernel":"causal_conv1d::causal_conv1d_update_l2norm_f32","family":"causal_conv1d_l2norm","point":null,"compute":"cuda_core","numerics":"reference","where":"device","state":"conv:snapshot"},{"id":"l33.gdn.l2","order":782,"layer":33,"block":"l33.gdn","block_type":"gdn","kind":"linear_attention","op":"l2_norm","roles":[],"inserted":false,"precision":{"weight":null,"activation":"f32","output":"f32"},"steps":["compute f32"],"group":473,"kernel":"causal_conv1d::causal_conv1d_update_l2norm_f32","family":"causal_conv1d_l2norm","point":null,"compute":"cuda_core","numerics":"reference","where":"device","state":null},{"id":"l33.gdn.recur","order":783,"layer":33,"block":"l33.gdn","block_type":"gdn","kind":"linear_attention","op":"gdn_recurrence","roles":[],"inserted":false,"precision":{"weight":null,"activation":"f32","output":"f32"},"steps":["state f32","compute f32"],"group":474,"kernel":"gated_delta_rule::gated_delta_rule_decode_f32","family":"gdn_recurrence","point":{"decay":"head","gate_act":"silu"},"compute":"cuda_core","numerics":"reference","where":"device","state":"h:update"},{"id":"l33.gdn.out_norm","order":784,"layer":33,"block":"l33.gdn","block_type":"gdn","kind":"linear_attention","op":"gated_rms_norm","roles":[],"inserted":false,"precision":{"weight":null,"activation":"f32","output":"bf16"},"steps":["compute f32"],"group":475,"kernel":"norm::gated_rms_norm_f32_input","family":"gated_rms_norm","point":null,"compute":"memory","numerics":"reference","where":"device","state":null},{"id":"l33.gdn.out","order":785,"layer":33,"block":"l33.gdn","block_type":"gdn","kind":"linear_attention","op":"linear:gdn_out","roles":["gdn_out"],"inserted":false,"precision":{"weight":"fp8/block128x128","activation":"bf16","output":"bf16"},"steps":["act bf16","weight fp8/block128x128->bf16","mma bf16*bf16","accumulate f32","scale f32"],"group":476,"kernel":"w8a16_gemm_pipelined_m32::w8a16_gemm_pipelined_m32","family":"w8a16_gemm","point":null,"compute":"tensor_core:mma.sync.m16n8k16.bf16","numerics":"reference","where":"device","state":null},{"id":"l33.gdn.add","order":786,"layer":33,"block":"l33.gdn","block_type":"gdn","kind":"linear_attention","op":"residual_add","roles":[],"inserted":false,"precision":{"weight":null,"activation":"bf16","output":"bf16"},"steps":["compute f32"],"group":477,"kernel":"norm::residual_add_rms_norm","family":"rms_norm","point":null,"compute":"memory","numerics":"reference","where":"device","state":null},{"id":"l33.moe_ffn.post_norm","order":787,"layer":33,"block":"l33.moe_ffn","block_type":"moe_ffn","kind":"linear_attention","op":"rms_norm","roles":[],"inserted":false,"precision":{"weight":null,"activation":"f32","output":"bf16"},"steps":["compute f32"],"group":477,"kernel":"norm::residual_add_rms_norm","family":"rms_norm","point":null,"compute":"memory","numerics":"reference","where":"device","state":null},{"id":"l33.moe_ffn.router","order":788,"layer":33,"block":"l33.moe_ffn","block_type":"moe_ffn","kind":"linear_attention","op":"router","roles":[],"inserted":false,"precision":{"weight":"bf16","activation":"bf16","output":"bf16"},"steps":["act bf16","weight bf16->bf16","mma bf16*bf16","accumulate f32","scale none"],"group":478,"kernel":"gemv::dense_gemv_bf16","family":"dense_bf16","point":null,"compute":"cuda_core","numerics":"reference","where":"device","state":null},{"id":"l33.moe_ffn.top_k","order":789,"layer":33,"block":"l33.moe_ffn","block_type":"moe_ffn","kind":"linear_attention","op":"top_k","roles":[],"inserted":false,"precision":{"weight":null,"activation":"bf16","output":"f32,i32"},"steps":["score f32"],"group":479,"kernel":"moe_topk::moe_topk_softmax","family":"moe_topk","point":null,"compute":"memory","numerics":"reference","where":"device","state":null},{"id":"l33.moe_ffn.experts_gate_up","order":790,"layer":33,"block":"l33.moe_ffn","block_type":"moe_ffn","kind":"linear_attention","op":"expert_gate_up","roles":[],"inserted":false,"precision":{"weight":"fp8/block128x128","activation":"bf16","output":"bf16"},"steps":["gather bf16","act bf16","weight fp8/block128x128->bf16","mma bf16*bf16","accumulate f32","scale f32"],"group":480,"kernel":"moe_shared_expert_fused_fp8::moe_expert_gate_up_shared_fp8","family":"moe_fp8_1row","point":null,"compute":"cuda_core","numerics":"reference","where":"device","state":null},{"id":"l33.moe_ffn.shared_gate_up","order":791,"layer":33,"block":"l33.moe_ffn","block_type":"moe_ffn","kind":"linear_attention","op":"linear:shared_gate_up","roles":["shared_gate_up"],"inserted":false,"precision":{"weight":"fp8/block128x128","activation":"bf16","output":"bf16"},"steps":["act bf16","weight fp8/block128x128->bf16","mma bf16*bf16","accumulate f32","scale f32"],"group":480,"kernel":"moe_shared_expert_fused_fp8::moe_expert_gate_up_shared_fp8","family":"moe_fp8_1row","point":null,"compute":"cuda_core","numerics":"reference","where":"device","state":null},{"id":"l33.moe_ffn.experts_act","order":792,"layer":33,"block":"l33.moe_ffn","block_type":"moe_ffn","kind":"linear_attention","op":"silu_mul","roles":[],"inserted":false,"precision":{"weight":null,"activation":"bf16","output":"f32"},"steps":["compute f32"],"group":481,"kernel":"moe_shared_expert_fused_fp8::moe_expert_silu_down_shared_fp8","family":"moe_fp8_1row","point":null,"compute":"cuda_core","numerics":"reference","where":"device","state":null},{"id":"l33.moe_ffn.experts_down","order":793,"layer":33,"block":"l33.moe_ffn","block_type":"moe_ffn","kind":"linear_attention","op":"expert_down","roles":[],"inserted":false,"precision":{"weight":"fp8/block128x128","activation":"f32","output":"bf16"},"steps":["act f32","weight fp8/block128x128->f32","mma f32*f32","accumulate f32","scale f32"],"group":481,"kernel":"moe_shared_expert_fused_fp8::moe_expert_silu_down_shared_fp8","family":"moe_fp8_1row","point":null,"compute":"cuda_core","numerics":"reference","where":"device","state":null},{"id":"l33.moe_ffn.shared_act","order":794,"layer":33,"block":"l33.moe_ffn","block_type":"moe_ffn","kind":"linear_attention","op":"silu_mul","roles":[],"inserted":false,"precision":{"weight":null,"activation":"bf16","output":"f32"},"steps":["compute f32"],"group":481,"kernel":"moe_shared_expert_fused_fp8::moe_expert_silu_down_shared_fp8","family":"moe_fp8_1row","point":null,"compute":"cuda_core","numerics":"reference","where":"device","state":null},{"id":"l33.moe_ffn.shared_down","order":795,"layer":33,"block":"l33.moe_ffn","block_type":"moe_ffn","kind":"linear_attention","op":"linear:shared_down","roles":["shared_down"],"inserted":false,"precision":{"weight":"fp8/block128x128","activation":"f32","output":"bf16"},"steps":["act f32","weight fp8/block128x128->f32","mma f32*f32","accumulate f32","scale f32"],"group":481,"kernel":"moe_shared_expert_fused_fp8::moe_expert_silu_down_shared_fp8","family":"moe_fp8_1row","point":null,"compute":"cuda_core","numerics":"reference","where":"device","state":null},{"id":"l33.moe_ffn.shared_gate","order":796,"layer":33,"block":"l33.moe_ffn","block_type":"moe_ffn","kind":"linear_attention","op":"linear:shared_gate","roles":["shared_gate"],"inserted":false,"precision":{"weight":"bf16","activation":"bf16","output":"f32"},"steps":["act bf16","weight bf16->bf16","mma bf16*bf16","accumulate f32","scale none"],"group":482,"kernel":"moe_expert_gemv::moe_weighted_sum_blend","family":"moe_blend","point":{"shared_gate":"sigmoid"},"compute":"cuda_core","numerics":"reference","where":"device","state":null},{"id":"l33.moe_ffn.blend","order":797,"layer":33,"block":"l33.moe_ffn","block_type":"moe_ffn","kind":"linear_attention","op":"blend","roles":[],"inserted":false,"precision":{"weight":null,"activation":"bf16","output":"bf16"},"steps":["scatter f32","combine f32"],"group":482,"kernel":"moe_expert_gemv::moe_weighted_sum_blend","family":"moe_blend","point":{"shared_gate":"sigmoid"},"compute":"cuda_core","numerics":"reference","where":"device","state":null},{"id":"l33.moe_ffn.add","order":798,"layer":33,"block":"l33.moe_ffn","block_type":"moe_ffn","kind":"linear_attention","op":"residual_add","roles":[],"inserted":false,"precision":{"weight":null,"activation":"bf16","output":"bf16"},"steps":["compute f32"],"group":483,"kernel":"residual_add_rms_norm_exact::residual_add_rms_norm_exact","family":"rms_norm","point":null,"compute":"memory","numerics":"bit_identical(residual_add_rms_norm_exact_microtest)","where":"device","state":null},{"id":"l34.gdn.input_norm","order":799,"layer":34,"block":"l34.gdn","block_type":"gdn","kind":"linear_attention","op":"rms_norm","roles":[],"inserted":false,"precision":{"weight":null,"activation":"bf16","output":"bf16"},"steps":["compute f32"],"group":483,"kernel":"residual_add_rms_norm_exact::residual_add_rms_norm_exact","family":"rms_norm","point":null,"compute":"memory","numerics":"bit_identical(residual_add_rms_norm_exact_microtest)","where":"device","state":null},{"id":"l34.gdn.qkvz","order":800,"layer":34,"block":"l34.gdn","block_type":"gdn","kind":"linear_attention","op":"linear:qkvz","roles":["qkvz"],"inserted":false,"precision":{"weight":"fp8/block128x128","activation":"bf16","output":"bf16"},"steps":["act bf16","weight fp8/block128x128->bf16","mma bf16*bf16","accumulate f32","scale f32"],"group":484,"kernel":"w8a16_gemm_pipelined_m32::w8a16_gemm_pipelined_m32","family":"w8a16_gemm","point":null,"compute":"tensor_core:mma.sync.m16n8k16.bf16","numerics":"reference","where":"device","state":null},{"id":"l34.gdn.ba","order":801,"layer":34,"block":"l34.gdn","block_type":"gdn","kind":"linear_attention","op":"linear:ba","roles":["ba"],"inserted":false,"precision":{"weight":"bf16","activation":"bf16","output":"f32"},"steps":["act bf16","weight bf16->bf16","mma bf16*bf16","accumulate f32","scale none"],"group":485,"kernel":"ssm_preprocess::dense_gemv_ba_gates","family":"gdn_ba_gates_gemv","point":{"gate":"softplus","decay":"head"},"compute":"cuda_core","numerics":"reference","where":"device","state":null},{"id":"l34.gdn.gates","order":802,"layer":34,"block":"l34.gdn","block_type":"gdn","kind":"linear_attention","op":"gdn_gates","roles":[],"inserted":false,"precision":{"weight":null,"activation":"f32","output":"f32,f32"},"steps":["compute f32"],"group":485,"kernel":"ssm_preprocess::dense_gemv_ba_gates","family":"gdn_ba_gates_gemv","point":{"gate":"softplus","decay":"head"},"compute":"cuda_core","numerics":"reference","where":"device","state":null},{"id":"l34.gdn.conv","order":803,"layer":34,"block":"l34.gdn","block_type":"gdn","kind":"linear_attention","op":"conv1d_update","roles":[],"inserted":false,"precision":{"weight":null,"activation":"bf16","output":"f32"},"steps":["state f32","compute f32"],"group":486,"kernel":"causal_conv1d::causal_conv1d_update_l2norm_f32","family":"causal_conv1d_l2norm","point":null,"compute":"cuda_core","numerics":"reference","where":"device","state":"conv:update"},{"id":"l34.gdn.conv_ckpt","order":804,"layer":34,"block":"l34.gdn","block_type":"gdn","kind":"linear_attention","op":"state_snapshot","roles":[],"inserted":false,"precision":{"weight":null,"activation":"f32","output":null},"steps":["state f32"],"group":486,"kernel":"causal_conv1d::causal_conv1d_update_l2norm_f32","family":"causal_conv1d_l2norm","point":null,"compute":"cuda_core","numerics":"reference","where":"device","state":"conv:snapshot"},{"id":"l34.gdn.l2","order":805,"layer":34,"block":"l34.gdn","block_type":"gdn","kind":"linear_attention","op":"l2_norm","roles":[],"inserted":false,"precision":{"weight":null,"activation":"f32","output":"f32"},"steps":["compute f32"],"group":486,"kernel":"causal_conv1d::causal_conv1d_update_l2norm_f32","family":"causal_conv1d_l2norm","point":null,"compute":"cuda_core","numerics":"reference","where":"device","state":null},{"id":"l34.gdn.recur","order":806,"layer":34,"block":"l34.gdn","block_type":"gdn","kind":"linear_attention","op":"gdn_recurrence","roles":[],"inserted":false,"precision":{"weight":null,"activation":"f32","output":"f32"},"steps":["state f32","compute f32"],"group":487,"kernel":"gated_delta_rule::gated_delta_rule_decode_f32","family":"gdn_recurrence","point":{"decay":"head","gate_act":"silu"},"compute":"cuda_core","numerics":"reference","where":"device","state":"h:update"},{"id":"l34.gdn.out_norm","order":807,"layer":34,"block":"l34.gdn","block_type":"gdn","kind":"linear_attention","op":"gated_rms_norm","roles":[],"inserted":false,"precision":{"weight":null,"activation":"f32","output":"bf16"},"steps":["compute f32"],"group":488,"kernel":"norm::gated_rms_norm_f32_input","family":"gated_rms_norm","point":null,"compute":"memory","numerics":"reference","where":"device","state":null},{"id":"l34.gdn.out","order":808,"layer":34,"block":"l34.gdn","block_type":"gdn","kind":"linear_attention","op":"linear:gdn_out","roles":["gdn_out"],"inserted":false,"precision":{"weight":"fp8/block128x128","activation":"bf16","output":"bf16"},"steps":["act bf16","weight fp8/block128x128->bf16","mma bf16*bf16","accumulate f32","scale f32"],"group":489,"kernel":"w8a16_gemm_pipelined_m32::w8a16_gemm_pipelined_m32","family":"w8a16_gemm","point":null,"compute":"tensor_core:mma.sync.m16n8k16.bf16","numerics":"reference","where":"device","state":null},{"id":"l34.gdn.add","order":809,"layer":34,"block":"l34.gdn","block_type":"gdn","kind":"linear_attention","op":"residual_add","roles":[],"inserted":false,"precision":{"weight":null,"activation":"bf16","output":"bf16"},"steps":["compute f32"],"group":490,"kernel":"norm::residual_add_rms_norm","family":"rms_norm","point":null,"compute":"memory","numerics":"reference","where":"device","state":null},{"id":"l34.moe_ffn.post_norm","order":810,"layer":34,"block":"l34.moe_ffn","block_type":"moe_ffn","kind":"linear_attention","op":"rms_norm","roles":[],"inserted":false,"precision":{"weight":null,"activation":"f32","output":"bf16"},"steps":["compute f32"],"group":490,"kernel":"norm::residual_add_rms_norm","family":"rms_norm","point":null,"compute":"memory","numerics":"reference","where":"device","state":null},{"id":"l34.moe_ffn.router","order":811,"layer":34,"block":"l34.moe_ffn","block_type":"moe_ffn","kind":"linear_attention","op":"router","roles":[],"inserted":false,"precision":{"weight":"bf16","activation":"bf16","output":"bf16"},"steps":["act bf16","weight bf16->bf16","mma bf16*bf16","accumulate f32","scale none"],"group":491,"kernel":"gemv::dense_gemv_bf16","family":"dense_bf16","point":null,"compute":"cuda_core","numerics":"reference","where":"device","state":null},{"id":"l34.moe_ffn.top_k","order":812,"layer":34,"block":"l34.moe_ffn","block_type":"moe_ffn","kind":"linear_attention","op":"top_k","roles":[],"inserted":false,"precision":{"weight":null,"activation":"bf16","output":"f32,i32"},"steps":["score f32"],"group":492,"kernel":"moe_topk::moe_topk_softmax","family":"moe_topk","point":null,"compute":"memory","numerics":"reference","where":"device","state":null},{"id":"l34.moe_ffn.experts_gate_up","order":813,"layer":34,"block":"l34.moe_ffn","block_type":"moe_ffn","kind":"linear_attention","op":"expert_gate_up","roles":[],"inserted":false,"precision":{"weight":"fp8/block128x128","activation":"bf16","output":"bf16"},"steps":["gather bf16","act bf16","weight fp8/block128x128->bf16","mma bf16*bf16","accumulate f32","scale f32"],"group":493,"kernel":"moe_shared_expert_fused_fp8::moe_expert_gate_up_shared_fp8","family":"moe_fp8_1row","point":null,"compute":"cuda_core","numerics":"reference","where":"device","state":null},{"id":"l34.moe_ffn.shared_gate_up","order":814,"layer":34,"block":"l34.moe_ffn","block_type":"moe_ffn","kind":"linear_attention","op":"linear:shared_gate_up","roles":["shared_gate_up"],"inserted":false,"precision":{"weight":"fp8/block128x128","activation":"bf16","output":"bf16"},"steps":["act bf16","weight fp8/block128x128->bf16","mma bf16*bf16","accumulate f32","scale f32"],"group":493,"kernel":"moe_shared_expert_fused_fp8::moe_expert_gate_up_shared_fp8","family":"moe_fp8_1row","point":null,"compute":"cuda_core","numerics":"reference","where":"device","state":null},{"id":"l34.moe_ffn.experts_act","order":815,"layer":34,"block":"l34.moe_ffn","block_type":"moe_ffn","kind":"linear_attention","op":"silu_mul","roles":[],"inserted":false,"precision":{"weight":null,"activation":"bf16","output":"f32"},"steps":["compute f32"],"group":494,"kernel":"moe_shared_expert_fused_fp8::moe_expert_silu_down_shared_fp8","family":"moe_fp8_1row","point":null,"compute":"cuda_core","numerics":"reference","where":"device","state":null},{"id":"l34.moe_ffn.experts_down","order":816,"layer":34,"block":"l34.moe_ffn","block_type":"moe_ffn","kind":"linear_attention","op":"expert_down","roles":[],"inserted":false,"precision":{"weight":"fp8/block128x128","activation":"f32","output":"bf16"},"steps":["act f32","weight fp8/block128x128->f32","mma f32*f32","accumulate f32","scale f32"],"group":494,"kernel":"moe_shared_expert_fused_fp8::moe_expert_silu_down_shared_fp8","family":"moe_fp8_1row","point":null,"compute":"cuda_core","numerics":"reference","where":"device","state":null},{"id":"l34.moe_ffn.shared_act","order":817,"layer":34,"block":"l34.moe_ffn","block_type":"moe_ffn","kind":"linear_attention","op":"silu_mul","roles":[],"inserted":false,"precision":{"weight":null,"activation":"bf16","output":"f32"},"steps":["compute f32"],"group":494,"kernel":"moe_shared_expert_fused_fp8::moe_expert_silu_down_shared_fp8","family":"moe_fp8_1row","point":null,"compute":"cuda_core","numerics":"reference","where":"device","state":null},{"id":"l34.moe_ffn.shared_down","order":818,"layer":34,"block":"l34.moe_ffn","block_type":"moe_ffn","kind":"linear_attention","op":"linear:shared_down","roles":["shared_down"],"inserted":false,"precision":{"weight":"fp8/block128x128","activation":"f32","output":"bf16"},"steps":["act f32","weight fp8/block128x128->f32","mma f32*f32","accumulate f32","scale f32"],"group":494,"kernel":"moe_shared_expert_fused_fp8::moe_expert_silu_down_shared_fp8","family":"moe_fp8_1row","point":null,"compute":"cuda_core","numerics":"reference","where":"device","state":null},{"id":"l34.moe_ffn.shared_gate","order":819,"layer":34,"block":"l34.moe_ffn","block_type":"moe_ffn","kind":"linear_attention","op":"linear:shared_gate","roles":["shared_gate"],"inserted":false,"precision":{"weight":"bf16","activation":"bf16","output":"f32"},"steps":["act bf16","weight bf16->bf16","mma bf16*bf16","accumulate f32","scale none"],"group":495,"kernel":"moe_expert_gemv::moe_weighted_sum_blend","family":"moe_blend","point":{"shared_gate":"sigmoid"},"compute":"cuda_core","numerics":"reference","where":"device","state":null},{"id":"l34.moe_ffn.blend","order":820,"layer":34,"block":"l34.moe_ffn","block_type":"moe_ffn","kind":"linear_attention","op":"blend","roles":[],"inserted":false,"precision":{"weight":null,"activation":"bf16","output":"bf16"},"steps":["scatter f32","combine f32"],"group":495,"kernel":"moe_expert_gemv::moe_weighted_sum_blend","family":"moe_blend","point":{"shared_gate":"sigmoid"},"compute":"cuda_core","numerics":"reference","where":"device","state":null},{"id":"l34.moe_ffn.add","order":821,"layer":34,"block":"l34.moe_ffn","block_type":"moe_ffn","kind":"linear_attention","op":"residual_add","roles":[],"inserted":false,"precision":{"weight":null,"activation":"bf16","output":"bf16"},"steps":["compute f32"],"group":496,"kernel":"residual_add_rms_norm_exact::residual_add_rms_norm_exact","family":"rms_norm","point":null,"compute":"memory","numerics":"bit_identical(residual_add_rms_norm_exact_microtest)","where":"device","state":null},{"id":"l35.attn.input_norm","order":822,"layer":35,"block":"l35.attn","block_type":"attn","kind":"full_attention","op":"rms_norm","roles":[],"inserted":false,"precision":{"weight":null,"activation":"bf16","output":"bf16"},"steps":["compute f32"],"group":496,"kernel":"residual_add_rms_norm_exact::residual_add_rms_norm_exact","family":"rms_norm","point":null,"compute":"memory","numerics":"bit_identical(residual_add_rms_norm_exact_microtest)","where":"device","state":null},{"id":"l35.attn.q","order":823,"layer":35,"block":"l35.attn","block_type":"attn","kind":"full_attention","op":"linear:q","roles":["q"],"inserted":false,"precision":{"weight":"fp8/block128x128","activation":"bf16","output":"bf16"},"steps":["act bf16","weight fp8/block128x128->bf16","mma bf16*bf16","accumulate f32","scale f32"],"group":497,"kernel":"w8a16_gemm_pipelined_m32::w8a16_gemm_pipelined_m32","family":"w8a16_gemm","point":null,"compute":"tensor_core:mma.sync.m16n8k16.bf16","numerics":"reference","where":"device","state":null},{"id":"l35.attn.q_split","order":824,"layer":35,"block":"l35.attn","block_type":"attn","kind":"full_attention","op":"split","roles":[],"inserted":false,"precision":{"weight":null,"activation":"bf16","output":"bf16,bf16"},"steps":["move bf16"],"group":498,"kernel":"ssm_preprocess::deinterleave_qg","family":"deinterleave_qg","point":{"layout":"interleaved"},"compute":"memory","numerics":"reference","where":"device","state":null},{"id":"l35.attn.k","order":825,"layer":35,"block":"l35.attn","block_type":"attn","kind":"full_attention","op":"linear:k","roles":["k"],"inserted":false,"precision":{"weight":"fp8/block128x128","activation":"bf16","output":"bf16"},"steps":["act bf16","weight fp8/block128x128->bf16","mma bf16*bf16","accumulate f32","scale f32"],"group":499,"kernel":"w8a16_gemm_pipelined_m32::w8a16_gemm_pipelined_m32","family":"w8a16_gemm","point":null,"compute":"tensor_core:mma.sync.m16n8k16.bf16","numerics":"reference","where":"device","state":null},{"id":"l35.attn.v","order":826,"layer":35,"block":"l35.attn","block_type":"attn","kind":"full_attention","op":"linear:v","roles":["v"],"inserted":false,"precision":{"weight":"fp8/block128x128","activation":"bf16","output":"bf16"},"steps":["act bf16","weight fp8/block128x128->bf16","mma bf16*bf16","accumulate f32","scale f32"],"group":500,"kernel":"w8a16_gemm_pipelined_m32::w8a16_gemm_pipelined_m32","family":"w8a16_gemm","point":null,"compute":"tensor_core:mma.sync.m16n8k16.bf16","numerics":"reference","where":"device","state":null},{"id":"l35.attn.q_norm","order":827,"layer":35,"block":"l35.attn","block_type":"attn","kind":"full_attention","op":"qk_norm","roles":[],"inserted":false,"precision":{"weight":null,"activation":"bf16","output":"bf16"},"steps":["compute f32"],"group":501,"kernel":"norm::rms_norm","family":"rms_norm","point":null,"compute":"memory","numerics":"reference","where":"device","state":null},{"id":"l35.attn.k_norm","order":828,"layer":35,"block":"l35.attn","block_type":"attn","kind":"full_attention","op":"qk_norm","roles":[],"inserted":false,"precision":{"weight":null,"activation":"bf16","output":"bf16"},"steps":["compute f32"],"group":502,"kernel":"norm::rms_norm","family":"rms_norm","point":null,"compute":"memory","numerics":"reference","where":"device","state":null},{"id":"l35.attn.rope","order":829,"layer":35,"block":"l35.attn","block_type":"attn","kind":"full_attention","op":"rope","roles":[],"inserted":false,"precision":{"weight":null,"activation":"bf16","output":"bf16,bf16"},"steps":["compute f32"],"group":503,"kernel":"rope_mrope_interleaved::rope_forward_mrope_interleaved","family":"rope","point":null,"compute":"memory","numerics":"reference","where":"device","state":null},{"id":"l35.attn.kv_write","order":830,"layer":35,"block":"l35.attn","block_type":"attn","kind":"full_attention","op":"kv_write","roles":[],"inserted":false,"precision":{"weight":null,"activation":"bf16","output":null},"steps":["cache bf16"],"group":504,"kernel":"reshape_and_cache::reshape_and_cache_flash","family":"kv_write","point":null,"compute":"memory","numerics":"reference","where":"device","state":"k:write,v:write"},{"id":"l35.attn.attend","order":831,"layer":35,"block":"l35.attn","block_type":"attn","kind":"full_attention","op":"paged_attention","roles":[],"inserted":false,"precision":{"weight":null,"activation":"bf16","output":"bf16"},"steps":["cache bf16","scores f32","softmax f32","accumulate f32"],"group":505,"kernel":"paged_decode::paged_decode_attn","family":"paged_decode_attn","point":null,"compute":"cuda_core","numerics":"reference","where":"device","state":"k:read,v:read"},{"id":"l35.attn.gate_mul","order":832,"layer":35,"block":"l35.attn","block_type":"attn","kind":"full_attention","op":"sigmoid_gate_mul","roles":[],"inserted":false,"precision":{"weight":null,"activation":"bf16","output":"bf16"},"steps":["compute f32"],"group":506,"kernel":"residual_add::sigmoid_gate_mul","family":"sigmoid_gate_mul","point":null,"compute":"memory","numerics":"reference","where":"device","state":null},{"id":"l35.attn.o","order":833,"layer":35,"block":"l35.attn","block_type":"attn","kind":"full_attention","op":"linear:o","roles":["o"],"inserted":false,"precision":{"weight":"fp8/block128x128","activation":"bf16","output":"bf16"},"steps":["act bf16","weight fp8/block128x128->bf16","mma bf16*bf16","accumulate f32","scale f32"],"group":507,"kernel":"w8a16_gemm_pipelined_m32::w8a16_gemm_pipelined_m32","family":"w8a16_gemm","point":null,"compute":"tensor_core:mma.sync.m16n8k16.bf16","numerics":"reference","where":"device","state":null},{"id":"l35.attn.add","order":834,"layer":35,"block":"l35.attn","block_type":"attn","kind":"full_attention","op":"residual_add","roles":[],"inserted":false,"precision":{"weight":null,"activation":"bf16","output":"bf16"},"steps":["compute f32"],"group":508,"kernel":"norm::residual_add_rms_norm","family":"rms_norm","point":null,"compute":"memory","numerics":"reference","where":"device","state":null},{"id":"l35.moe_ffn.post_norm","order":835,"layer":35,"block":"l35.moe_ffn","block_type":"moe_ffn","kind":"full_attention","op":"rms_norm","roles":[],"inserted":false,"precision":{"weight":null,"activation":"f32","output":"bf16"},"steps":["compute f32"],"group":508,"kernel":"norm::residual_add_rms_norm","family":"rms_norm","point":null,"compute":"memory","numerics":"reference","where":"device","state":null},{"id":"l35.moe_ffn.router","order":836,"layer":35,"block":"l35.moe_ffn","block_type":"moe_ffn","kind":"full_attention","op":"router","roles":[],"inserted":false,"precision":{"weight":"bf16","activation":"bf16","output":"bf16"},"steps":["act bf16","weight bf16->bf16","mma bf16*bf16","accumulate f32","scale none"],"group":509,"kernel":"gemv::dense_gemv_bf16","family":"dense_bf16","point":null,"compute":"cuda_core","numerics":"reference","where":"device","state":null},{"id":"l35.moe_ffn.top_k","order":837,"layer":35,"block":"l35.moe_ffn","block_type":"moe_ffn","kind":"full_attention","op":"top_k","roles":[],"inserted":false,"precision":{"weight":null,"activation":"bf16","output":"f32,i32"},"steps":["score f32"],"group":510,"kernel":"moe_topk::moe_topk_softmax","family":"moe_topk","point":null,"compute":"memory","numerics":"reference","where":"device","state":null},{"id":"l35.moe_ffn.experts_gate_up","order":838,"layer":35,"block":"l35.moe_ffn","block_type":"moe_ffn","kind":"full_attention","op":"expert_gate_up","roles":[],"inserted":false,"precision":{"weight":"fp8/block128x128","activation":"bf16","output":"bf16"},"steps":["gather bf16","act bf16","weight fp8/block128x128->bf16","mma bf16*bf16","accumulate f32","scale f32"],"group":511,"kernel":"moe_shared_expert_fused_fp8::moe_expert_gate_up_shared_fp8","family":"moe_fp8_1row","point":null,"compute":"cuda_core","numerics":"reference","where":"device","state":null},{"id":"l35.moe_ffn.shared_gate_up","order":839,"layer":35,"block":"l35.moe_ffn","block_type":"moe_ffn","kind":"full_attention","op":"linear:shared_gate_up","roles":["shared_gate_up"],"inserted":false,"precision":{"weight":"fp8/block128x128","activation":"bf16","output":"bf16"},"steps":["act bf16","weight fp8/block128x128->bf16","mma bf16*bf16","accumulate f32","scale f32"],"group":511,"kernel":"moe_shared_expert_fused_fp8::moe_expert_gate_up_shared_fp8","family":"moe_fp8_1row","point":null,"compute":"cuda_core","numerics":"reference","where":"device","state":null},{"id":"l35.moe_ffn.experts_act","order":840,"layer":35,"block":"l35.moe_ffn","block_type":"moe_ffn","kind":"full_attention","op":"silu_mul","roles":[],"inserted":false,"precision":{"weight":null,"activation":"bf16","output":"f32"},"steps":["compute f32"],"group":512,"kernel":"moe_shared_expert_fused_fp8::moe_expert_silu_down_shared_fp8","family":"moe_fp8_1row","point":null,"compute":"cuda_core","numerics":"reference","where":"device","state":null},{"id":"l35.moe_ffn.experts_down","order":841,"layer":35,"block":"l35.moe_ffn","block_type":"moe_ffn","kind":"full_attention","op":"expert_down","roles":[],"inserted":false,"precision":{"weight":"fp8/block128x128","activation":"f32","output":"bf16"},"steps":["act f32","weight fp8/block128x128->f32","mma f32*f32","accumulate f32","scale f32"],"group":512,"kernel":"moe_shared_expert_fused_fp8::moe_expert_silu_down_shared_fp8","family":"moe_fp8_1row","point":null,"compute":"cuda_core","numerics":"reference","where":"device","state":null},{"id":"l35.moe_ffn.shared_act","order":842,"layer":35,"block":"l35.moe_ffn","block_type":"moe_ffn","kind":"full_attention","op":"silu_mul","roles":[],"inserted":false,"precision":{"weight":null,"activation":"bf16","output":"f32"},"steps":["compute f32"],"group":512,"kernel":"moe_shared_expert_fused_fp8::moe_expert_silu_down_shared_fp8","family":"moe_fp8_1row","point":null,"compute":"cuda_core","numerics":"reference","where":"device","state":null},{"id":"l35.moe_ffn.shared_down","order":843,"layer":35,"block":"l35.moe_ffn","block_type":"moe_ffn","kind":"full_attention","op":"linear:shared_down","roles":["shared_down"],"inserted":false,"precision":{"weight":"fp8/block128x128","activation":"f32","output":"bf16"},"steps":["act f32","weight fp8/block128x128->f32","mma f32*f32","accumulate f32","scale f32"],"group":512,"kernel":"moe_shared_expert_fused_fp8::moe_expert_silu_down_shared_fp8","family":"moe_fp8_1row","point":null,"compute":"cuda_core","numerics":"reference","where":"device","state":null},{"id":"l35.moe_ffn.shared_gate","order":844,"layer":35,"block":"l35.moe_ffn","block_type":"moe_ffn","kind":"full_attention","op":"linear:shared_gate","roles":["shared_gate"],"inserted":false,"precision":{"weight":"bf16","activation":"bf16","output":"f32"},"steps":["act bf16","weight bf16->bf16","mma bf16*bf16","accumulate f32","scale none"],"group":513,"kernel":"moe_expert_gemv::moe_weighted_sum_blend","family":"moe_blend","point":{"shared_gate":"sigmoid"},"compute":"cuda_core","numerics":"reference","where":"device","state":null},{"id":"l35.moe_ffn.blend","order":845,"layer":35,"block":"l35.moe_ffn","block_type":"moe_ffn","kind":"full_attention","op":"blend","roles":[],"inserted":false,"precision":{"weight":null,"activation":"bf16","output":"bf16"},"steps":["scatter f32","combine f32"],"group":513,"kernel":"moe_expert_gemv::moe_weighted_sum_blend","family":"moe_blend","point":{"shared_gate":"sigmoid"},"compute":"cuda_core","numerics":"reference","where":"device","state":null},{"id":"l35.moe_ffn.add","order":846,"layer":35,"block":"l35.moe_ffn","block_type":"moe_ffn","kind":"full_attention","op":"residual_add","roles":[],"inserted":false,"precision":{"weight":null,"activation":"bf16","output":"bf16"},"steps":["compute f32"],"group":514,"kernel":"residual_add_rms_norm_exact::residual_add_rms_norm_exact","family":"rms_norm","point":null,"compute":"memory","numerics":"bit_identical(residual_add_rms_norm_exact_microtest)","where":"device","state":null},{"id":"l36.gdn.input_norm","order":847,"layer":36,"block":"l36.gdn","block_type":"gdn","kind":"linear_attention","op":"rms_norm","roles":[],"inserted":false,"precision":{"weight":null,"activation":"bf16","output":"bf16"},"steps":["compute f32"],"group":514,"kernel":"residual_add_rms_norm_exact::residual_add_rms_norm_exact","family":"rms_norm","point":null,"compute":"memory","numerics":"bit_identical(residual_add_rms_norm_exact_microtest)","where":"device","state":null},{"id":"l36.gdn.qkvz","order":848,"layer":36,"block":"l36.gdn","block_type":"gdn","kind":"linear_attention","op":"linear:qkvz","roles":["qkvz"],"inserted":false,"precision":{"weight":"fp8/block128x128","activation":"bf16","output":"bf16"},"steps":["act bf16","weight fp8/block128x128->bf16","mma bf16*bf16","accumulate f32","scale f32"],"group":515,"kernel":"w8a16_gemm_pipelined_m32::w8a16_gemm_pipelined_m32","family":"w8a16_gemm","point":null,"compute":"tensor_core:mma.sync.m16n8k16.bf16","numerics":"reference","where":"device","state":null},{"id":"l36.gdn.ba","order":849,"layer":36,"block":"l36.gdn","block_type":"gdn","kind":"linear_attention","op":"linear:ba","roles":["ba"],"inserted":false,"precision":{"weight":"bf16","activation":"bf16","output":"f32"},"steps":["act bf16","weight bf16->bf16","mma bf16*bf16","accumulate f32","scale none"],"group":516,"kernel":"ssm_preprocess::dense_gemv_ba_gates","family":"gdn_ba_gates_gemv","point":{"gate":"softplus","decay":"head"},"compute":"cuda_core","numerics":"reference","where":"device","state":null},{"id":"l36.gdn.gates","order":850,"layer":36,"block":"l36.gdn","block_type":"gdn","kind":"linear_attention","op":"gdn_gates","roles":[],"inserted":false,"precision":{"weight":null,"activation":"f32","output":"f32,f32"},"steps":["compute f32"],"group":516,"kernel":"ssm_preprocess::dense_gemv_ba_gates","family":"gdn_ba_gates_gemv","point":{"gate":"softplus","decay":"head"},"compute":"cuda_core","numerics":"reference","where":"device","state":null},{"id":"l36.gdn.conv","order":851,"layer":36,"block":"l36.gdn","block_type":"gdn","kind":"linear_attention","op":"conv1d_update","roles":[],"inserted":false,"precision":{"weight":null,"activation":"bf16","output":"f32"},"steps":["state f32","compute f32"],"group":517,"kernel":"causal_conv1d::causal_conv1d_update_l2norm_f32","family":"causal_conv1d_l2norm","point":null,"compute":"cuda_core","numerics":"reference","where":"device","state":"conv:update"},{"id":"l36.gdn.conv_ckpt","order":852,"layer":36,"block":"l36.gdn","block_type":"gdn","kind":"linear_attention","op":"state_snapshot","roles":[],"inserted":false,"precision":{"weight":null,"activation":"f32","output":null},"steps":["state f32"],"group":517,"kernel":"causal_conv1d::causal_conv1d_update_l2norm_f32","family":"causal_conv1d_l2norm","point":null,"compute":"cuda_core","numerics":"reference","where":"device","state":"conv:snapshot"},{"id":"l36.gdn.l2","order":853,"layer":36,"block":"l36.gdn","block_type":"gdn","kind":"linear_attention","op":"l2_norm","roles":[],"inserted":false,"precision":{"weight":null,"activation":"f32","output":"f32"},"steps":["compute f32"],"group":517,"kernel":"causal_conv1d::causal_conv1d_update_l2norm_f32","family":"causal_conv1d_l2norm","point":null,"compute":"cuda_core","numerics":"reference","where":"device","state":null},{"id":"l36.gdn.recur","order":854,"layer":36,"block":"l36.gdn","block_type":"gdn","kind":"linear_attention","op":"gdn_recurrence","roles":[],"inserted":false,"precision":{"weight":null,"activation":"f32","output":"f32"},"steps":["state f32","compute f32"],"group":518,"kernel":"gated_delta_rule::gated_delta_rule_decode_f32","family":"gdn_recurrence","point":{"decay":"head","gate_act":"silu"},"compute":"cuda_core","numerics":"reference","where":"device","state":"h:update"},{"id":"l36.gdn.out_norm","order":855,"layer":36,"block":"l36.gdn","block_type":"gdn","kind":"linear_attention","op":"gated_rms_norm","roles":[],"inserted":false,"precision":{"weight":null,"activation":"f32","output":"bf16"},"steps":["compute f32"],"group":519,"kernel":"norm::gated_rms_norm_f32_input","family":"gated_rms_norm","point":null,"compute":"memory","numerics":"reference","where":"device","state":null},{"id":"l36.gdn.out","order":856,"layer":36,"block":"l36.gdn","block_type":"gdn","kind":"linear_attention","op":"linear:gdn_out","roles":["gdn_out"],"inserted":false,"precision":{"weight":"fp8/block128x128","activation":"bf16","output":"bf16"},"steps":["act bf16","weight fp8/block128x128->bf16","mma bf16*bf16","accumulate f32","scale f32"],"group":520,"kernel":"w8a16_gemm_pipelined_m32::w8a16_gemm_pipelined_m32","family":"w8a16_gemm","point":null,"compute":"tensor_core:mma.sync.m16n8k16.bf16","numerics":"reference","where":"device","state":null},{"id":"l36.gdn.add","order":857,"layer":36,"block":"l36.gdn","block_type":"gdn","kind":"linear_attention","op":"residual_add","roles":[],"inserted":false,"precision":{"weight":null,"activation":"bf16","output":"bf16"},"steps":["compute f32"],"group":521,"kernel":"norm::residual_add_rms_norm","family":"rms_norm","point":null,"compute":"memory","numerics":"reference","where":"device","state":null},{"id":"l36.moe_ffn.post_norm","order":858,"layer":36,"block":"l36.moe_ffn","block_type":"moe_ffn","kind":"linear_attention","op":"rms_norm","roles":[],"inserted":false,"precision":{"weight":null,"activation":"f32","output":"bf16"},"steps":["compute f32"],"group":521,"kernel":"norm::residual_add_rms_norm","family":"rms_norm","point":null,"compute":"memory","numerics":"reference","where":"device","state":null},{"id":"l36.moe_ffn.router","order":859,"layer":36,"block":"l36.moe_ffn","block_type":"moe_ffn","kind":"linear_attention","op":"router","roles":[],"inserted":false,"precision":{"weight":"bf16","activation":"bf16","output":"bf16"},"steps":["act bf16","weight bf16->bf16","mma bf16*bf16","accumulate f32","scale none"],"group":522,"kernel":"gemv::dense_gemv_bf16","family":"dense_bf16","point":null,"compute":"cuda_core","numerics":"reference","where":"device","state":null},{"id":"l36.moe_ffn.top_k","order":860,"layer":36,"block":"l36.moe_ffn","block_type":"moe_ffn","kind":"linear_attention","op":"top_k","roles":[],"inserted":false,"precision":{"weight":null,"activation":"bf16","output":"f32,i32"},"steps":["score f32"],"group":523,"kernel":"moe_topk::moe_topk_softmax","family":"moe_topk","point":null,"compute":"memory","numerics":"reference","where":"device","state":null},{"id":"l36.moe_ffn.experts_gate_up","order":861,"layer":36,"block":"l36.moe_ffn","block_type":"moe_ffn","kind":"linear_attention","op":"expert_gate_up","roles":[],"inserted":false,"precision":{"weight":"fp8/block128x128","activation":"bf16","output":"bf16"},"steps":["gather bf16","act bf16","weight fp8/block128x128->bf16","mma bf16*bf16","accumulate f32","scale f32"],"group":524,"kernel":"moe_shared_expert_fused_fp8::moe_expert_gate_up_shared_fp8","family":"moe_fp8_1row","point":null,"compute":"cuda_core","numerics":"reference","where":"device","state":null},{"id":"l36.moe_ffn.shared_gate_up","order":862,"layer":36,"block":"l36.moe_ffn","block_type":"moe_ffn","kind":"linear_attention","op":"linear:shared_gate_up","roles":["shared_gate_up"],"inserted":false,"precision":{"weight":"fp8/block128x128","activation":"bf16","output":"bf16"},"steps":["act bf16","weight fp8/block128x128->bf16","mma bf16*bf16","accumulate f32","scale f32"],"group":524,"kernel":"moe_shared_expert_fused_fp8::moe_expert_gate_up_shared_fp8","family":"moe_fp8_1row","point":null,"compute":"cuda_core","numerics":"reference","where":"device","state":null},{"id":"l36.moe_ffn.experts_act","order":863,"layer":36,"block":"l36.moe_ffn","block_type":"moe_ffn","kind":"linear_attention","op":"silu_mul","roles":[],"inserted":false,"precision":{"weight":null,"activation":"bf16","output":"f32"},"steps":["compute f32"],"group":525,"kernel":"moe_shared_expert_fused_fp8::moe_expert_silu_down_shared_fp8","family":"moe_fp8_1row","point":null,"compute":"cuda_core","numerics":"reference","where":"device","state":null},{"id":"l36.moe_ffn.experts_down","order":864,"layer":36,"block":"l36.moe_ffn","block_type":"moe_ffn","kind":"linear_attention","op":"expert_down","roles":[],"inserted":false,"precision":{"weight":"fp8/block128x128","activation":"f32","output":"bf16"},"steps":["act f32","weight fp8/block128x128->f32","mma f32*f32","accumulate f32","scale f32"],"group":525,"kernel":"moe_shared_expert_fused_fp8::moe_expert_silu_down_shared_fp8","family":"moe_fp8_1row","point":null,"compute":"cuda_core","numerics":"reference","where":"device","state":null},{"id":"l36.moe_ffn.shared_act","order":865,"layer":36,"block":"l36.moe_ffn","block_type":"moe_ffn","kind":"linear_attention","op":"silu_mul","roles":[],"inserted":false,"precision":{"weight":null,"activation":"bf16","output":"f32"},"steps":["compute f32"],"group":525,"kernel":"moe_shared_expert_fused_fp8::moe_expert_silu_down_shared_fp8","family":"moe_fp8_1row","point":null,"compute":"cuda_core","numerics":"reference","where":"device","state":null},{"id":"l36.moe_ffn.shared_down","order":866,"layer":36,"block":"l36.moe_ffn","block_type":"moe_ffn","kind":"linear_attention","op":"linear:shared_down","roles":["shared_down"],"inserted":false,"precision":{"weight":"fp8/block128x128","activation":"f32","output":"bf16"},"steps":["act f32","weight fp8/block128x128->f32","mma f32*f32","accumulate f32","scale f32"],"group":525,"kernel":"moe_shared_expert_fused_fp8::moe_expert_silu_down_shared_fp8","family":"moe_fp8_1row","point":null,"compute":"cuda_core","numerics":"reference","where":"device","state":null},{"id":"l36.moe_ffn.shared_gate","order":867,"layer":36,"block":"l36.moe_ffn","block_type":"moe_ffn","kind":"linear_attention","op":"linear:shared_gate","roles":["shared_gate"],"inserted":false,"precision":{"weight":"bf16","activation":"bf16","output":"f32"},"steps":["act bf16","weight bf16->bf16","mma bf16*bf16","accumulate f32","scale none"],"group":526,"kernel":"moe_expert_gemv::moe_weighted_sum_blend","family":"moe_blend","point":{"shared_gate":"sigmoid"},"compute":"cuda_core","numerics":"reference","where":"device","state":null},{"id":"l36.moe_ffn.blend","order":868,"layer":36,"block":"l36.moe_ffn","block_type":"moe_ffn","kind":"linear_attention","op":"blend","roles":[],"inserted":false,"precision":{"weight":null,"activation":"bf16","output":"bf16"},"steps":["scatter f32","combine f32"],"group":526,"kernel":"moe_expert_gemv::moe_weighted_sum_blend","family":"moe_blend","point":{"shared_gate":"sigmoid"},"compute":"cuda_core","numerics":"reference","where":"device","state":null},{"id":"l36.moe_ffn.add","order":869,"layer":36,"block":"l36.moe_ffn","block_type":"moe_ffn","kind":"linear_attention","op":"residual_add","roles":[],"inserted":false,"precision":{"weight":null,"activation":"bf16","output":"bf16"},"steps":["compute f32"],"group":527,"kernel":"residual_add_rms_norm_exact::residual_add_rms_norm_exact","family":"rms_norm","point":null,"compute":"memory","numerics":"bit_identical(residual_add_rms_norm_exact_microtest)","where":"device","state":null},{"id":"l37.gdn.input_norm","order":870,"layer":37,"block":"l37.gdn","block_type":"gdn","kind":"linear_attention","op":"rms_norm","roles":[],"inserted":false,"precision":{"weight":null,"activation":"bf16","output":"bf16"},"steps":["compute f32"],"group":527,"kernel":"residual_add_rms_norm_exact::residual_add_rms_norm_exact","family":"rms_norm","point":null,"compute":"memory","numerics":"bit_identical(residual_add_rms_norm_exact_microtest)","where":"device","state":null},{"id":"l37.gdn.qkvz","order":871,"layer":37,"block":"l37.gdn","block_type":"gdn","kind":"linear_attention","op":"linear:qkvz","roles":["qkvz"],"inserted":false,"precision":{"weight":"fp8/block128x128","activation":"bf16","output":"bf16"},"steps":["act bf16","weight fp8/block128x128->bf16","mma bf16*bf16","accumulate f32","scale f32"],"group":528,"kernel":"w8a16_gemm_pipelined_m32::w8a16_gemm_pipelined_m32","family":"w8a16_gemm","point":null,"compute":"tensor_core:mma.sync.m16n8k16.bf16","numerics":"reference","where":"device","state":null},{"id":"l37.gdn.ba","order":872,"layer":37,"block":"l37.gdn","block_type":"gdn","kind":"linear_attention","op":"linear:ba","roles":["ba"],"inserted":false,"precision":{"weight":"bf16","activation":"bf16","output":"f32"},"steps":["act bf16","weight bf16->bf16","mma bf16*bf16","accumulate f32","scale none"],"group":529,"kernel":"ssm_preprocess::dense_gemv_ba_gates","family":"gdn_ba_gates_gemv","point":{"gate":"softplus","decay":"head"},"compute":"cuda_core","numerics":"reference","where":"device","state":null},{"id":"l37.gdn.gates","order":873,"layer":37,"block":"l37.gdn","block_type":"gdn","kind":"linear_attention","op":"gdn_gates","roles":[],"inserted":false,"precision":{"weight":null,"activation":"f32","output":"f32,f32"},"steps":["compute f32"],"group":529,"kernel":"ssm_preprocess::dense_gemv_ba_gates","family":"gdn_ba_gates_gemv","point":{"gate":"softplus","decay":"head"},"compute":"cuda_core","numerics":"reference","where":"device","state":null},{"id":"l37.gdn.conv","order":874,"layer":37,"block":"l37.gdn","block_type":"gdn","kind":"linear_attention","op":"conv1d_update","roles":[],"inserted":false,"precision":{"weight":null,"activation":"bf16","output":"f32"},"steps":["state f32","compute f32"],"group":530,"kernel":"causal_conv1d::causal_conv1d_update_l2norm_f32","family":"causal_conv1d_l2norm","point":null,"compute":"cuda_core","numerics":"reference","where":"device","state":"conv:update"},{"id":"l37.gdn.conv_ckpt","order":875,"layer":37,"block":"l37.gdn","block_type":"gdn","kind":"linear_attention","op":"state_snapshot","roles":[],"inserted":false,"precision":{"weight":null,"activation":"f32","output":null},"steps":["state f32"],"group":530,"kernel":"causal_conv1d::causal_conv1d_update_l2norm_f32","family":"causal_conv1d_l2norm","point":null,"compute":"cuda_core","numerics":"reference","where":"device","state":"conv:snapshot"},{"id":"l37.gdn.l2","order":876,"layer":37,"block":"l37.gdn","block_type":"gdn","kind":"linear_attention","op":"l2_norm","roles":[],"inserted":false,"precision":{"weight":null,"activation":"f32","output":"f32"},"steps":["compute f32"],"group":530,"kernel":"causal_conv1d::causal_conv1d_update_l2norm_f32","family":"causal_conv1d_l2norm","point":null,"compute":"cuda_core","numerics":"reference","where":"device","state":null},{"id":"l37.gdn.recur","order":877,"layer":37,"block":"l37.gdn","block_type":"gdn","kind":"linear_attention","op":"gdn_recurrence","roles":[],"inserted":false,"precision":{"weight":null,"activation":"f32","output":"f32"},"steps":["state f32","compute f32"],"group":531,"kernel":"gated_delta_rule::gated_delta_rule_decode_f32","family":"gdn_recurrence","point":{"decay":"head","gate_act":"silu"},"compute":"cuda_core","numerics":"reference","where":"device","state":"h:update"},{"id":"l37.gdn.out_norm","order":878,"layer":37,"block":"l37.gdn","block_type":"gdn","kind":"linear_attention","op":"gated_rms_norm","roles":[],"inserted":false,"precision":{"weight":null,"activation":"f32","output":"bf16"},"steps":["compute f32"],"group":532,"kernel":"norm::gated_rms_norm_f32_input","family":"gated_rms_norm","point":null,"compute":"memory","numerics":"reference","where":"device","state":null},{"id":"l37.gdn.out","order":879,"layer":37,"block":"l37.gdn","block_type":"gdn","kind":"linear_attention","op":"linear:gdn_out","roles":["gdn_out"],"inserted":false,"precision":{"weight":"fp8/block128x128","activation":"bf16","output":"bf16"},"steps":["act bf16","weight fp8/block128x128->bf16","mma bf16*bf16","accumulate f32","scale f32"],"group":533,"kernel":"w8a16_gemm_pipelined_m32::w8a16_gemm_pipelined_m32","family":"w8a16_gemm","point":null,"compute":"tensor_core:mma.sync.m16n8k16.bf16","numerics":"reference","where":"device","state":null},{"id":"l37.gdn.add","order":880,"layer":37,"block":"l37.gdn","block_type":"gdn","kind":"linear_attention","op":"residual_add","roles":[],"inserted":false,"precision":{"weight":null,"activation":"bf16","output":"bf16"},"steps":["compute f32"],"group":534,"kernel":"norm::residual_add_rms_norm","family":"rms_norm","point":null,"compute":"memory","numerics":"reference","where":"device","state":null},{"id":"l37.moe_ffn.post_norm","order":881,"layer":37,"block":"l37.moe_ffn","block_type":"moe_ffn","kind":"linear_attention","op":"rms_norm","roles":[],"inserted":false,"precision":{"weight":null,"activation":"f32","output":"bf16"},"steps":["compute f32"],"group":534,"kernel":"norm::residual_add_rms_norm","family":"rms_norm","point":null,"compute":"memory","numerics":"reference","where":"device","state":null},{"id":"l37.moe_ffn.router","order":882,"layer":37,"block":"l37.moe_ffn","block_type":"moe_ffn","kind":"linear_attention","op":"router","roles":[],"inserted":false,"precision":{"weight":"bf16","activation":"bf16","output":"bf16"},"steps":["act bf16","weight bf16->bf16","mma bf16*bf16","accumulate f32","scale none"],"group":535,"kernel":"gemv::dense_gemv_bf16","family":"dense_bf16","point":null,"compute":"cuda_core","numerics":"reference","where":"device","state":null},{"id":"l37.moe_ffn.top_k","order":883,"layer":37,"block":"l37.moe_ffn","block_type":"moe_ffn","kind":"linear_attention","op":"top_k","roles":[],"inserted":false,"precision":{"weight":null,"activation":"bf16","output":"f32,i32"},"steps":["score f32"],"group":536,"kernel":"moe_topk::moe_topk_softmax","family":"moe_topk","point":null,"compute":"memory","numerics":"reference","where":"device","state":null},{"id":"l37.moe_ffn.experts_gate_up","order":884,"layer":37,"block":"l37.moe_ffn","block_type":"moe_ffn","kind":"linear_attention","op":"expert_gate_up","roles":[],"inserted":false,"precision":{"weight":"fp8/block128x128","activation":"bf16","output":"bf16"},"steps":["gather bf16","act bf16","weight fp8/block128x128->bf16","mma bf16*bf16","accumulate f32","scale f32"],"group":537,"kernel":"moe_shared_expert_fused_fp8::moe_expert_gate_up_shared_fp8","family":"moe_fp8_1row","point":null,"compute":"cuda_core","numerics":"reference","where":"device","state":null},{"id":"l37.moe_ffn.shared_gate_up","order":885,"layer":37,"block":"l37.moe_ffn","block_type":"moe_ffn","kind":"linear_attention","op":"linear:shared_gate_up","roles":["shared_gate_up"],"inserted":false,"precision":{"weight":"fp8/block128x128","activation":"bf16","output":"bf16"},"steps":["act bf16","weight fp8/block128x128->bf16","mma bf16*bf16","accumulate f32","scale f32"],"group":537,"kernel":"moe_shared_expert_fused_fp8::moe_expert_gate_up_shared_fp8","family":"moe_fp8_1row","point":null,"compute":"cuda_core","numerics":"reference","where":"device","state":null},{"id":"l37.moe_ffn.experts_act","order":886,"layer":37,"block":"l37.moe_ffn","block_type":"moe_ffn","kind":"linear_attention","op":"silu_mul","roles":[],"inserted":false,"precision":{"weight":null,"activation":"bf16","output":"f32"},"steps":["compute f32"],"group":538,"kernel":"moe_shared_expert_fused_fp8::moe_expert_silu_down_shared_fp8","family":"moe_fp8_1row","point":null,"compute":"cuda_core","numerics":"reference","where":"device","state":null},{"id":"l37.moe_ffn.experts_down","order":887,"layer":37,"block":"l37.moe_ffn","block_type":"moe_ffn","kind":"linear_attention","op":"expert_down","roles":[],"inserted":false,"precision":{"weight":"fp8/block128x128","activation":"f32","output":"bf16"},"steps":["act f32","weight fp8/block128x128->f32","mma f32*f32","accumulate f32","scale f32"],"group":538,"kernel":"moe_shared_expert_fused_fp8::moe_expert_silu_down_shared_fp8","family":"moe_fp8_1row","point":null,"compute":"cuda_core","numerics":"reference","where":"device","state":null},{"id":"l37.moe_ffn.shared_act","order":888,"layer":37,"block":"l37.moe_ffn","block_type":"moe_ffn","kind":"linear_attention","op":"silu_mul","roles":[],"inserted":false,"precision":{"weight":null,"activation":"bf16","output":"f32"},"steps":["compute f32"],"group":538,"kernel":"moe_shared_expert_fused_fp8::moe_expert_silu_down_shared_fp8","family":"moe_fp8_1row","point":null,"compute":"cuda_core","numerics":"reference","where":"device","state":null},{"id":"l37.moe_ffn.shared_down","order":889,"layer":37,"block":"l37.moe_ffn","block_type":"moe_ffn","kind":"linear_attention","op":"linear:shared_down","roles":["shared_down"],"inserted":false,"precision":{"weight":"fp8/block128x128","activation":"f32","output":"bf16"},"steps":["act f32","weight fp8/block128x128->f32","mma f32*f32","accumulate f32","scale f32"],"group":538,"kernel":"moe_shared_expert_fused_fp8::moe_expert_silu_down_shared_fp8","family":"moe_fp8_1row","point":null,"compute":"cuda_core","numerics":"reference","where":"device","state":null},{"id":"l37.moe_ffn.shared_gate","order":890,"layer":37,"block":"l37.moe_ffn","block_type":"moe_ffn","kind":"linear_attention","op":"linear:shared_gate","roles":["shared_gate"],"inserted":false,"precision":{"weight":"bf16","activation":"bf16","output":"f32"},"steps":["act bf16","weight bf16->bf16","mma bf16*bf16","accumulate f32","scale none"],"group":539,"kernel":"moe_expert_gemv::moe_weighted_sum_blend","family":"moe_blend","point":{"shared_gate":"sigmoid"},"compute":"cuda_core","numerics":"reference","where":"device","state":null},{"id":"l37.moe_ffn.blend","order":891,"layer":37,"block":"l37.moe_ffn","block_type":"moe_ffn","kind":"linear_attention","op":"blend","roles":[],"inserted":false,"precision":{"weight":null,"activation":"bf16","output":"bf16"},"steps":["scatter f32","combine f32"],"group":539,"kernel":"moe_expert_gemv::moe_weighted_sum_blend","family":"moe_blend","point":{"shared_gate":"sigmoid"},"compute":"cuda_core","numerics":"reference","where":"device","state":null},{"id":"l37.moe_ffn.add","order":892,"layer":37,"block":"l37.moe_ffn","block_type":"moe_ffn","kind":"linear_attention","op":"residual_add","roles":[],"inserted":false,"precision":{"weight":null,"activation":"bf16","output":"bf16"},"steps":["compute f32"],"group":540,"kernel":"residual_add_rms_norm_exact::residual_add_rms_norm_exact","family":"rms_norm","point":null,"compute":"memory","numerics":"bit_identical(residual_add_rms_norm_exact_microtest)","where":"device","state":null},{"id":"l38.gdn.input_norm","order":893,"layer":38,"block":"l38.gdn","block_type":"gdn","kind":"linear_attention","op":"rms_norm","roles":[],"inserted":false,"precision":{"weight":null,"activation":"bf16","output":"bf16"},"steps":["compute f32"],"group":540,"kernel":"residual_add_rms_norm_exact::residual_add_rms_norm_exact","family":"rms_norm","point":null,"compute":"memory","numerics":"bit_identical(residual_add_rms_norm_exact_microtest)","where":"device","state":null},{"id":"l38.gdn.qkvz","order":894,"layer":38,"block":"l38.gdn","block_type":"gdn","kind":"linear_attention","op":"linear:qkvz","roles":["qkvz"],"inserted":false,"precision":{"weight":"fp8/block128x128","activation":"bf16","output":"bf16"},"steps":["act bf16","weight fp8/block128x128->bf16","mma bf16*bf16","accumulate f32","scale f32"],"group":541,"kernel":"w8a16_gemm_pipelined_m32::w8a16_gemm_pipelined_m32","family":"w8a16_gemm","point":null,"compute":"tensor_core:mma.sync.m16n8k16.bf16","numerics":"reference","where":"device","state":null},{"id":"l38.gdn.ba","order":895,"layer":38,"block":"l38.gdn","block_type":"gdn","kind":"linear_attention","op":"linear:ba","roles":["ba"],"inserted":false,"precision":{"weight":"bf16","activation":"bf16","output":"f32"},"steps":["act bf16","weight bf16->bf16","mma bf16*bf16","accumulate f32","scale none"],"group":542,"kernel":"ssm_preprocess::dense_gemv_ba_gates","family":"gdn_ba_gates_gemv","point":{"gate":"softplus","decay":"head"},"compute":"cuda_core","numerics":"reference","where":"device","state":null},{"id":"l38.gdn.gates","order":896,"layer":38,"block":"l38.gdn","block_type":"gdn","kind":"linear_attention","op":"gdn_gates","roles":[],"inserted":false,"precision":{"weight":null,"activation":"f32","output":"f32,f32"},"steps":["compute f32"],"group":542,"kernel":"ssm_preprocess::dense_gemv_ba_gates","family":"gdn_ba_gates_gemv","point":{"gate":"softplus","decay":"head"},"compute":"cuda_core","numerics":"reference","where":"device","state":null},{"id":"l38.gdn.conv","order":897,"layer":38,"block":"l38.gdn","block_type":"gdn","kind":"linear_attention","op":"conv1d_update","roles":[],"inserted":false,"precision":{"weight":null,"activation":"bf16","output":"f32"},"steps":["state f32","compute f32"],"group":543,"kernel":"causal_conv1d::causal_conv1d_update_l2norm_f32","family":"causal_conv1d_l2norm","point":null,"compute":"cuda_core","numerics":"reference","where":"device","state":"conv:update"},{"id":"l38.gdn.conv_ckpt","order":898,"layer":38,"block":"l38.gdn","block_type":"gdn","kind":"linear_attention","op":"state_snapshot","roles":[],"inserted":false,"precision":{"weight":null,"activation":"f32","output":null},"steps":["state f32"],"group":543,"kernel":"causal_conv1d::causal_conv1d_update_l2norm_f32","family":"causal_conv1d_l2norm","point":null,"compute":"cuda_core","numerics":"reference","where":"device","state":"conv:snapshot"},{"id":"l38.gdn.l2","order":899,"layer":38,"block":"l38.gdn","block_type":"gdn","kind":"linear_attention","op":"l2_norm","roles":[],"inserted":false,"precision":{"weight":null,"activation":"f32","output":"f32"},"steps":["compute f32"],"group":543,"kernel":"causal_conv1d::causal_conv1d_update_l2norm_f32","family":"causal_conv1d_l2norm","point":null,"compute":"cuda_core","numerics":"reference","where":"device","state":null},{"id":"l38.gdn.recur","order":900,"layer":38,"block":"l38.gdn","block_type":"gdn","kind":"linear_attention","op":"gdn_recurrence","roles":[],"inserted":false,"precision":{"weight":null,"activation":"f32","output":"f32"},"steps":["state f32","compute f32"],"group":544,"kernel":"gated_delta_rule::gated_delta_rule_decode_f32","family":"gdn_recurrence","point":{"decay":"head","gate_act":"silu"},"compute":"cuda_core","numerics":"reference","where":"device","state":"h:update"},{"id":"l38.gdn.out_norm","order":901,"layer":38,"block":"l38.gdn","block_type":"gdn","kind":"linear_attention","op":"gated_rms_norm","roles":[],"inserted":false,"precision":{"weight":null,"activation":"f32","output":"bf16"},"steps":["compute f32"],"group":545,"kernel":"norm::gated_rms_norm_f32_input","family":"gated_rms_norm","point":null,"compute":"memory","numerics":"reference","where":"device","state":null},{"id":"l38.gdn.out","order":902,"layer":38,"block":"l38.gdn","block_type":"gdn","kind":"linear_attention","op":"linear:gdn_out","roles":["gdn_out"],"inserted":false,"precision":{"weight":"fp8/block128x128","activation":"bf16","output":"bf16"},"steps":["act bf16","weight fp8/block128x128->bf16","mma bf16*bf16","accumulate f32","scale f32"],"group":546,"kernel":"w8a16_gemm_pipelined_m32::w8a16_gemm_pipelined_m32","family":"w8a16_gemm","point":null,"compute":"tensor_core:mma.sync.m16n8k16.bf16","numerics":"reference","where":"device","state":null},{"id":"l38.gdn.add","order":903,"layer":38,"block":"l38.gdn","block_type":"gdn","kind":"linear_attention","op":"residual_add","roles":[],"inserted":false,"precision":{"weight":null,"activation":"bf16","output":"bf16"},"steps":["compute f32"],"group":547,"kernel":"norm::residual_add_rms_norm","family":"rms_norm","point":null,"compute":"memory","numerics":"reference","where":"device","state":null},{"id":"l38.moe_ffn.post_norm","order":904,"layer":38,"block":"l38.moe_ffn","block_type":"moe_ffn","kind":"linear_attention","op":"rms_norm","roles":[],"inserted":false,"precision":{"weight":null,"activation":"f32","output":"bf16"},"steps":["compute f32"],"group":547,"kernel":"norm::residual_add_rms_norm","family":"rms_norm","point":null,"compute":"memory","numerics":"reference","where":"device","state":null},{"id":"l38.moe_ffn.router","order":905,"layer":38,"block":"l38.moe_ffn","block_type":"moe_ffn","kind":"linear_attention","op":"router","roles":[],"inserted":false,"precision":{"weight":"bf16","activation":"bf16","output":"bf16"},"steps":["act bf16","weight bf16->bf16","mma bf16*bf16","accumulate f32","scale none"],"group":548,"kernel":"gemv::dense_gemv_bf16","family":"dense_bf16","point":null,"compute":"cuda_core","numerics":"reference","where":"device","state":null},{"id":"l38.moe_ffn.top_k","order":906,"layer":38,"block":"l38.moe_ffn","block_type":"moe_ffn","kind":"linear_attention","op":"top_k","roles":[],"inserted":false,"precision":{"weight":null,"activation":"bf16","output":"f32,i32"},"steps":["score f32"],"group":549,"kernel":"moe_topk::moe_topk_softmax","family":"moe_topk","point":null,"compute":"memory","numerics":"reference","where":"device","state":null},{"id":"l38.moe_ffn.experts_gate_up","order":907,"layer":38,"block":"l38.moe_ffn","block_type":"moe_ffn","kind":"linear_attention","op":"expert_gate_up","roles":[],"inserted":false,"precision":{"weight":"fp8/block128x128","activation":"bf16","output":"bf16"},"steps":["gather bf16","act bf16","weight fp8/block128x128->bf16","mma bf16*bf16","accumulate f32","scale f32"],"group":550,"kernel":"moe_shared_expert_fused_fp8::moe_expert_gate_up_shared_fp8","family":"moe_fp8_1row","point":null,"compute":"cuda_core","numerics":"reference","where":"device","state":null},{"id":"l38.moe_ffn.shared_gate_up","order":908,"layer":38,"block":"l38.moe_ffn","block_type":"moe_ffn","kind":"linear_attention","op":"linear:shared_gate_up","roles":["shared_gate_up"],"inserted":false,"precision":{"weight":"fp8/block128x128","activation":"bf16","output":"bf16"},"steps":["act bf16","weight fp8/block128x128->bf16","mma bf16*bf16","accumulate f32","scale f32"],"group":550,"kernel":"moe_shared_expert_fused_fp8::moe_expert_gate_up_shared_fp8","family":"moe_fp8_1row","point":null,"compute":"cuda_core","numerics":"reference","where":"device","state":null},{"id":"l38.moe_ffn.experts_act","order":909,"layer":38,"block":"l38.moe_ffn","block_type":"moe_ffn","kind":"linear_attention","op":"silu_mul","roles":[],"inserted":false,"precision":{"weight":null,"activation":"bf16","output":"f32"},"steps":["compute f32"],"group":551,"kernel":"moe_shared_expert_fused_fp8::moe_expert_silu_down_shared_fp8","family":"moe_fp8_1row","point":null,"compute":"cuda_core","numerics":"reference","where":"device","state":null},{"id":"l38.moe_ffn.experts_down","order":910,"layer":38,"block":"l38.moe_ffn","block_type":"moe_ffn","kind":"linear_attention","op":"expert_down","roles":[],"inserted":false,"precision":{"weight":"fp8/block128x128","activation":"f32","output":"bf16"},"steps":["act f32","weight fp8/block128x128->f32","mma f32*f32","accumulate f32","scale f32"],"group":551,"kernel":"moe_shared_expert_fused_fp8::moe_expert_silu_down_shared_fp8","family":"moe_fp8_1row","point":null,"compute":"cuda_core","numerics":"reference","where":"device","state":null},{"id":"l38.moe_ffn.shared_act","order":911,"layer":38,"block":"l38.moe_ffn","block_type":"moe_ffn","kind":"linear_attention","op":"silu_mul","roles":[],"inserted":false,"precision":{"weight":null,"activation":"bf16","output":"f32"},"steps":["compute f32"],"group":551,"kernel":"moe_shared_expert_fused_fp8::moe_expert_silu_down_shared_fp8","family":"moe_fp8_1row","point":null,"compute":"cuda_core","numerics":"reference","where":"device","state":null},{"id":"l38.moe_ffn.shared_down","order":912,"layer":38,"block":"l38.moe_ffn","block_type":"moe_ffn","kind":"linear_attention","op":"linear:shared_down","roles":["shared_down"],"inserted":false,"precision":{"weight":"fp8/block128x128","activation":"f32","output":"bf16"},"steps":["act f32","weight fp8/block128x128->f32","mma f32*f32","accumulate f32","scale f32"],"group":551,"kernel":"moe_shared_expert_fused_fp8::moe_expert_silu_down_shared_fp8","family":"moe_fp8_1row","point":null,"compute":"cuda_core","numerics":"reference","where":"device","state":null},{"id":"l38.moe_ffn.shared_gate","order":913,"layer":38,"block":"l38.moe_ffn","block_type":"moe_ffn","kind":"linear_attention","op":"linear:shared_gate","roles":["shared_gate"],"inserted":false,"precision":{"weight":"bf16","activation":"bf16","output":"f32"},"steps":["act bf16","weight bf16->bf16","mma bf16*bf16","accumulate f32","scale none"],"group":552,"kernel":"moe_expert_gemv::moe_weighted_sum_blend","family":"moe_blend","point":{"shared_gate":"sigmoid"},"compute":"cuda_core","numerics":"reference","where":"device","state":null},{"id":"l38.moe_ffn.blend","order":914,"layer":38,"block":"l38.moe_ffn","block_type":"moe_ffn","kind":"linear_attention","op":"blend","roles":[],"inserted":false,"precision":{"weight":null,"activation":"bf16","output":"bf16"},"steps":["scatter f32","combine f32"],"group":552,"kernel":"moe_expert_gemv::moe_weighted_sum_blend","family":"moe_blend","point":{"shared_gate":"sigmoid"},"compute":"cuda_core","numerics":"reference","where":"device","state":null},{"id":"l38.moe_ffn.add","order":915,"layer":38,"block":"l38.moe_ffn","block_type":"moe_ffn","kind":"linear_attention","op":"residual_add","roles":[],"inserted":false,"precision":{"weight":null,"activation":"bf16","output":"bf16"},"steps":["compute f32"],"group":553,"kernel":"residual_add_rms_norm_exact::residual_add_rms_norm_exact","family":"rms_norm","point":null,"compute":"memory","numerics":"bit_identical(residual_add_rms_norm_exact_microtest)","where":"device","state":null},{"id":"l39.attn.input_norm","order":916,"layer":39,"block":"l39.attn","block_type":"attn","kind":"full_attention","op":"rms_norm","roles":[],"inserted":false,"precision":{"weight":null,"activation":"bf16","output":"bf16"},"steps":["compute f32"],"group":553,"kernel":"residual_add_rms_norm_exact::residual_add_rms_norm_exact","family":"rms_norm","point":null,"compute":"memory","numerics":"bit_identical(residual_add_rms_norm_exact_microtest)","where":"device","state":null},{"id":"l39.attn.q","order":917,"layer":39,"block":"l39.attn","block_type":"attn","kind":"full_attention","op":"linear:q","roles":["q"],"inserted":false,"precision":{"weight":"fp8/block128x128","activation":"bf16","output":"bf16"},"steps":["act bf16","weight fp8/block128x128->bf16","mma bf16*bf16","accumulate f32","scale f32"],"group":554,"kernel":"w8a16_gemm_pipelined_m32::w8a16_gemm_pipelined_m32","family":"w8a16_gemm","point":null,"compute":"tensor_core:mma.sync.m16n8k16.bf16","numerics":"reference","where":"device","state":null},{"id":"l39.attn.q_split","order":918,"layer":39,"block":"l39.attn","block_type":"attn","kind":"full_attention","op":"split","roles":[],"inserted":false,"precision":{"weight":null,"activation":"bf16","output":"bf16,bf16"},"steps":["move bf16"],"group":555,"kernel":"ssm_preprocess::deinterleave_qg","family":"deinterleave_qg","point":{"layout":"interleaved"},"compute":"memory","numerics":"reference","where":"device","state":null},{"id":"l39.attn.k","order":919,"layer":39,"block":"l39.attn","block_type":"attn","kind":"full_attention","op":"linear:k","roles":["k"],"inserted":false,"precision":{"weight":"fp8/block128x128","activation":"bf16","output":"bf16"},"steps":["act bf16","weight fp8/block128x128->bf16","mma bf16*bf16","accumulate f32","scale f32"],"group":556,"kernel":"w8a16_gemm_pipelined_m32::w8a16_gemm_pipelined_m32","family":"w8a16_gemm","point":null,"compute":"tensor_core:mma.sync.m16n8k16.bf16","numerics":"reference","where":"device","state":null},{"id":"l39.attn.v","order":920,"layer":39,"block":"l39.attn","block_type":"attn","kind":"full_attention","op":"linear:v","roles":["v"],"inserted":false,"precision":{"weight":"fp8/block128x128","activation":"bf16","output":"bf16"},"steps":["act bf16","weight fp8/block128x128->bf16","mma bf16*bf16","accumulate f32","scale f32"],"group":557,"kernel":"w8a16_gemm_pipelined_m32::w8a16_gemm_pipelined_m32","family":"w8a16_gemm","point":null,"compute":"tensor_core:mma.sync.m16n8k16.bf16","numerics":"reference","where":"device","state":null},{"id":"l39.attn.q_norm","order":921,"layer":39,"block":"l39.attn","block_type":"attn","kind":"full_attention","op":"qk_norm","roles":[],"inserted":false,"precision":{"weight":null,"activation":"bf16","output":"bf16"},"steps":["compute f32"],"group":558,"kernel":"norm::rms_norm","family":"rms_norm","point":null,"compute":"memory","numerics":"reference","where":"device","state":null},{"id":"l39.attn.k_norm","order":922,"layer":39,"block":"l39.attn","block_type":"attn","kind":"full_attention","op":"qk_norm","roles":[],"inserted":false,"precision":{"weight":null,"activation":"bf16","output":"bf16"},"steps":["compute f32"],"group":559,"kernel":"norm::rms_norm","family":"rms_norm","point":null,"compute":"memory","numerics":"reference","where":"device","state":null},{"id":"l39.attn.rope","order":923,"layer":39,"block":"l39.attn","block_type":"attn","kind":"full_attention","op":"rope","roles":[],"inserted":false,"precision":{"weight":null,"activation":"bf16","output":"bf16,bf16"},"steps":["compute f32"],"group":560,"kernel":"rope_mrope_interleaved::rope_forward_mrope_interleaved","family":"rope","point":null,"compute":"memory","numerics":"reference","where":"device","state":null},{"id":"l39.attn.kv_write","order":924,"layer":39,"block":"l39.attn","block_type":"attn","kind":"full_attention","op":"kv_write","roles":[],"inserted":false,"precision":{"weight":null,"activation":"bf16","output":null},"steps":["cache bf16"],"group":561,"kernel":"reshape_and_cache::reshape_and_cache_flash","family":"kv_write","point":null,"compute":"memory","numerics":"reference","where":"device","state":"k:write,v:write"},{"id":"l39.attn.attend","order":925,"layer":39,"block":"l39.attn","block_type":"attn","kind":"full_attention","op":"paged_attention","roles":[],"inserted":false,"precision":{"weight":null,"activation":"bf16","output":"bf16"},"steps":["cache bf16","scores f32","softmax f32","accumulate f32"],"group":562,"kernel":"paged_decode::paged_decode_attn","family":"paged_decode_attn","point":null,"compute":"cuda_core","numerics":"reference","where":"device","state":"k:read,v:read"},{"id":"l39.attn.gate_mul","order":926,"layer":39,"block":"l39.attn","block_type":"attn","kind":"full_attention","op":"sigmoid_gate_mul","roles":[],"inserted":false,"precision":{"weight":null,"activation":"bf16","output":"bf16"},"steps":["compute f32"],"group":563,"kernel":"residual_add::sigmoid_gate_mul","family":"sigmoid_gate_mul","point":null,"compute":"memory","numerics":"reference","where":"device","state":null},{"id":"l39.attn.o","order":927,"layer":39,"block":"l39.attn","block_type":"attn","kind":"full_attention","op":"linear:o","roles":["o"],"inserted":false,"precision":{"weight":"fp8/block128x128","activation":"bf16","output":"bf16"},"steps":["act bf16","weight fp8/block128x128->bf16","mma bf16*bf16","accumulate f32","scale f32"],"group":564,"kernel":"w8a16_gemm_pipelined_m32::w8a16_gemm_pipelined_m32","family":"w8a16_gemm","point":null,"compute":"tensor_core:mma.sync.m16n8k16.bf16","numerics":"reference","where":"device","state":null},{"id":"l39.attn.add","order":928,"layer":39,"block":"l39.attn","block_type":"attn","kind":"full_attention","op":"residual_add","roles":[],"inserted":false,"precision":{"weight":null,"activation":"bf16","output":"bf16"},"steps":["compute f32"],"group":565,"kernel":"norm::residual_add_rms_norm","family":"rms_norm","point":null,"compute":"memory","numerics":"reference","where":"device","state":null},{"id":"l39.moe_ffn.post_norm","order":929,"layer":39,"block":"l39.moe_ffn","block_type":"moe_ffn","kind":"full_attention","op":"rms_norm","roles":[],"inserted":false,"precision":{"weight":null,"activation":"f32","output":"bf16"},"steps":["compute f32"],"group":565,"kernel":"norm::residual_add_rms_norm","family":"rms_norm","point":null,"compute":"memory","numerics":"reference","where":"device","state":null},{"id":"l39.moe_ffn.router","order":930,"layer":39,"block":"l39.moe_ffn","block_type":"moe_ffn","kind":"full_attention","op":"router","roles":[],"inserted":false,"precision":{"weight":"bf16","activation":"bf16","output":"bf16"},"steps":["act bf16","weight bf16->bf16","mma bf16*bf16","accumulate f32","scale none"],"group":566,"kernel":"gemv::dense_gemv_bf16","family":"dense_bf16","point":null,"compute":"cuda_core","numerics":"reference","where":"device","state":null},{"id":"l39.moe_ffn.top_k","order":931,"layer":39,"block":"l39.moe_ffn","block_type":"moe_ffn","kind":"full_attention","op":"top_k","roles":[],"inserted":false,"precision":{"weight":null,"activation":"bf16","output":"f32,i32"},"steps":["score f32"],"group":567,"kernel":"moe_topk::moe_topk_softmax","family":"moe_topk","point":null,"compute":"memory","numerics":"reference","where":"device","state":null},{"id":"l39.moe_ffn.experts_gate_up","order":932,"layer":39,"block":"l39.moe_ffn","block_type":"moe_ffn","kind":"full_attention","op":"expert_gate_up","roles":[],"inserted":false,"precision":{"weight":"fp8/block128x128","activation":"bf16","output":"bf16"},"steps":["gather bf16","act bf16","weight fp8/block128x128->bf16","mma bf16*bf16","accumulate f32","scale f32"],"group":568,"kernel":"moe_shared_expert_fused_fp8::moe_expert_gate_up_shared_fp8","family":"moe_fp8_1row","point":null,"compute":"cuda_core","numerics":"reference","where":"device","state":null},{"id":"l39.moe_ffn.shared_gate_up","order":933,"layer":39,"block":"l39.moe_ffn","block_type":"moe_ffn","kind":"full_attention","op":"linear:shared_gate_up","roles":["shared_gate_up"],"inserted":false,"precision":{"weight":"fp8/block128x128","activation":"bf16","output":"bf16"},"steps":["act bf16","weight fp8/block128x128->bf16","mma bf16*bf16","accumulate f32","scale f32"],"group":568,"kernel":"moe_shared_expert_fused_fp8::moe_expert_gate_up_shared_fp8","family":"moe_fp8_1row","point":null,"compute":"cuda_core","numerics":"reference","where":"device","state":null},{"id":"l39.moe_ffn.experts_act","order":934,"layer":39,"block":"l39.moe_ffn","block_type":"moe_ffn","kind":"full_attention","op":"silu_mul","roles":[],"inserted":false,"precision":{"weight":null,"activation":"bf16","output":"f32"},"steps":["compute f32"],"group":569,"kernel":"moe_shared_expert_fused_fp8::moe_expert_silu_down_shared_fp8","family":"moe_fp8_1row","point":null,"compute":"cuda_core","numerics":"reference","where":"device","state":null},{"id":"l39.moe_ffn.experts_down","order":935,"layer":39,"block":"l39.moe_ffn","block_type":"moe_ffn","kind":"full_attention","op":"expert_down","roles":[],"inserted":false,"precision":{"weight":"fp8/block128x128","activation":"f32","output":"bf16"},"steps":["act f32","weight fp8/block128x128->f32","mma f32*f32","accumulate f32","scale f32"],"group":569,"kernel":"moe_shared_expert_fused_fp8::moe_expert_silu_down_shared_fp8","family":"moe_fp8_1row","point":null,"compute":"cuda_core","numerics":"reference","where":"device","state":null},{"id":"l39.moe_ffn.shared_act","order":936,"layer":39,"block":"l39.moe_ffn","block_type":"moe_ffn","kind":"full_attention","op":"silu_mul","roles":[],"inserted":false,"precision":{"weight":null,"activation":"bf16","output":"f32"},"steps":["compute f32"],"group":569,"kernel":"moe_shared_expert_fused_fp8::moe_expert_silu_down_shared_fp8","family":"moe_fp8_1row","point":null,"compute":"cuda_core","numerics":"reference","where":"device","state":null},{"id":"l39.moe_ffn.shared_down","order":937,"layer":39,"block":"l39.moe_ffn","block_type":"moe_ffn","kind":"full_attention","op":"linear:shared_down","roles":["shared_down"],"inserted":false,"precision":{"weight":"fp8/block128x128","activation":"f32","output":"bf16"},"steps":["act f32","weight fp8/block128x128->f32","mma f32*f32","accumulate f32","scale f32"],"group":569,"kernel":"moe_shared_expert_fused_fp8::moe_expert_silu_down_shared_fp8","family":"moe_fp8_1row","point":null,"compute":"cuda_core","numerics":"reference","where":"device","state":null},{"id":"l39.moe_ffn.shared_gate","order":938,"layer":39,"block":"l39.moe_ffn","block_type":"moe_ffn","kind":"full_attention","op":"linear:shared_gate","roles":["shared_gate"],"inserted":false,"precision":{"weight":"bf16","activation":"bf16","output":"f32"},"steps":["act bf16","weight bf16->bf16","mma bf16*bf16","accumulate f32","scale none"],"group":570,"kernel":"moe_expert_gemv::moe_weighted_sum_blend","family":"moe_blend","point":{"shared_gate":"sigmoid"},"compute":"cuda_core","numerics":"reference","where":"device","state":null},{"id":"l39.moe_ffn.blend","order":939,"layer":39,"block":"l39.moe_ffn","block_type":"moe_ffn","kind":"full_attention","op":"blend","roles":[],"inserted":false,"precision":{"weight":null,"activation":"bf16","output":"bf16"},"steps":["scatter f32","combine f32"],"group":570,"kernel":"moe_expert_gemv::moe_weighted_sum_blend","family":"moe_blend","point":{"shared_gate":"sigmoid"},"compute":"cuda_core","numerics":"reference","where":"device","state":null},{"id":"l39.moe_ffn.add","order":940,"layer":39,"block":"l39.moe_ffn","block_type":"moe_ffn","kind":"full_attention","op":"residual_add","roles":[],"inserted":false,"precision":{"weight":null,"activation":"bf16","output":"bf16"},"steps":["compute f32"],"group":571,"kernel":"residual_add::bf16_residual_add","family":"residual_add","point":null,"compute":"memory","numerics":"reference","where":"device","state":null},{"id":"head.final_norm","order":941,"layer":null,"block":"head","block_type":"head","kind":null,"op":"final_norm","roles":[],"inserted":false,"precision":{"weight":null,"activation":"bf16","output":"bf16"},"steps":["compute f32"],"group":572,"kernel":"norm::rms_norm","family":"rms_norm","point":null,"compute":"memory","numerics":"reference","where":"device","state":null},{"id":"head.lm_head","order":942,"layer":null,"block":"head","block_type":"head","kind":null,"op":"lm_head","roles":[],"inserted":false,"precision":{"weight":"bf16","activation":"bf16","output":"bf16"},"steps":["act bf16","weight bf16->bf16","mma bf16*bf16","accumulate f32","scale none"],"group":573,"kernel":"gemv::dense_gemv_bf16","family":"dense_bf16","point":null,"compute":"cuda_core","numerics":"reference","where":"device","state":null},{"id":"head.argmax","order":943,"layer":null,"block":"head","block_type":"head","kind":null,"op":"argmax","roles":[],"inserted":false,"precision":{"weight":null,"activation":"bf16","output":"i32"},"steps":["compare bf16"],"group":574,"kernel":null,"family":null,"point":null,"compute":null,"numerics":"reference","where":"host","state":null}],"edges":[{"id":"embed.h","from":0,"to":1,"format":"bf16","rows":"n","dim":"hidden","fused":false},{"id":"embed.h","from":0,"to":11,"format":"bf16","rows":"n","dim":"hidden","fused":false},{"id":"l0.gdn.xn","from":1,"to":2,"format":"bf16","rows":"n","dim":"hidden","fused":false},{"id":"l0.gdn.xn","from":1,"to":3,"format":"bf16","rows":"n","dim":"hidden","fused":false},{"id":"l0.gdn.qkvz","from":2,"to":5,"format":"bf16","rows":"n","dim":"lin_k_heads*lin_k_dim*2+lin_v_heads*lin_v_dim*2","fused":false},{"id":"l0.gdn.qkvz","from":2,"to":9,"format":"bf16","rows":"n","dim":"lin_k_heads*lin_k_dim*2+lin_v_heads*lin_v_dim*2","fused":false},{"id":"l0.gdn.ba","from":3,"to":4,"format":"f32","rows":"n","dim":"lin_v_heads*2","fused":true},{"id":"l0.gdn.decay","from":4,"to":8,"format":"f32","rows":"n","dim":"lin_v_heads","fused":false},{"id":"l0.gdn.beta","from":4,"to":8,"format":"f32","rows":"n","dim":"lin_v_heads","fused":false},{"id":"l0.gdn.conv","from":5,"to":6,"format":"f32","rows":"n","dim":"lin_k_heads*lin_k_dim*2+lin_v_heads*lin_v_dim","fused":true},{"id":"l0.gdn.conv","from":5,"to":7,"format":"f32","rows":"n","dim":"lin_k_heads*lin_k_dim*2+lin_v_heads*lin_v_dim","fused":true},{"id":"l0.gdn.qkv","from":7,"to":8,"format":"f32","rows":"n","dim":"lin_k_heads*lin_k_dim*2+lin_v_heads*lin_v_dim","fused":false},{"id":"l0.gdn.core","from":8,"to":9,"format":"f32","rows":"n","dim":"lin_v_heads*lin_v_dim","fused":false},{"id":"l0.gdn.gated","from":9,"to":10,"format":"bf16","rows":"n","dim":"lin_v_heads*lin_v_dim","fused":false},{"id":"l0.gdn.o","from":10,"to":11,"format":"bf16","rows":"n","dim":"hidden","fused":false},{"id":"l0.gdn.y","from":11,"to":12,"format":"bf16","rows":"n","dim":"hidden","fused":false},{"id":"l0.gdn.y","from":11,"to":23,"format":"bf16","rows":"n","dim":"hidden","fused":false},{"id":"l0.moe_ffn.xn","from":12,"to":13,"format":"bf16","rows":"n","dim":"hidden","fused":false},{"id":"l0.moe_ffn.xn","from":12,"to":15,"format":"bf16","rows":"n","dim":"hidden","fused":false},{"id":"l0.moe_ffn.xn","from":12,"to":16,"format":"bf16","rows":"n","dim":"hidden","fused":false},{"id":"l0.moe_ffn.xn","from":12,"to":21,"format":"bf16","rows":"n","dim":"hidden","fused":false},{"id":"l0.moe_ffn.router_logits","from":13,"to":14,"format":"bf16","rows":"n","dim":"experts","fused":false},{"id":"l0.moe_ffn.topk_w","from":14,"to":22,"format":"f32","rows":"n","dim":"top_k","fused":false},{"id":"l0.moe_ffn.topk_id","from":14,"to":15,"format":"i32","rows":"n","dim":"top_k","fused":false},{"id":"l0.moe_ffn.topk_id","from":14,"to":18,"format":"i32","rows":"n","dim":"top_k","fused":false},{"id":"l0.moe_ffn.egu","from":15,"to":17,"format":"bf16","rows":"n*top_k","dim":"moe_inter*2","fused":false},{"id":"l0.moe_ffn.sgu","from":16,"to":19,"format":"bf16","rows":"n","dim":"shared_inter*2","fused":false},{"id":"l0.moe_ffn.eact","from":17,"to":18,"format":"f32","rows":"n*top_k","dim":"moe_inter","fused":true},{"id":"l0.moe_ffn.edown","from":18,"to":22,"format":"bf16","rows":"n*top_k","dim":"hidden","fused":false},{"id":"l0.moe_ffn.sact","from":19,"to":20,"format":"f32","rows":"n","dim":"shared_inter","fused":true},{"id":"l0.moe_ffn.sdown","from":20,"to":22,"format":"bf16","rows":"n","dim":"hidden","fused":false},{"id":"l0.moe_ffn.sg","from":21,"to":22,"format":"f32","rows":"n","dim":"1","fused":true},{"id":"l0.moe_ffn.f","from":22,"to":23,"format":"bf16","rows":"n","dim":"hidden","fused":false},{"id":"l0.moe_ffn.y","from":23,"to":24,"format":"bf16","rows":"n","dim":"hidden","fused":false},{"id":"l0.moe_ffn.y","from":23,"to":34,"format":"bf16","rows":"n","dim":"hidden","fused":false},{"id":"l1.gdn.xn","from":24,"to":25,"format":"bf16","rows":"n","dim":"hidden","fused":false},{"id":"l1.gdn.xn","from":24,"to":26,"format":"bf16","rows":"n","dim":"hidden","fused":false},{"id":"l1.gdn.qkvz","from":25,"to":28,"format":"bf16","rows":"n","dim":"lin_k_heads*lin_k_dim*2+lin_v_heads*lin_v_dim*2","fused":false},{"id":"l1.gdn.qkvz","from":25,"to":32,"format":"bf16","rows":"n","dim":"lin_k_heads*lin_k_dim*2+lin_v_heads*lin_v_dim*2","fused":false},{"id":"l1.gdn.ba","from":26,"to":27,"format":"f32","rows":"n","dim":"lin_v_heads*2","fused":true},{"id":"l1.gdn.decay","from":27,"to":31,"format":"f32","rows":"n","dim":"lin_v_heads","fused":false},{"id":"l1.gdn.beta","from":27,"to":31,"format":"f32","rows":"n","dim":"lin_v_heads","fused":false},{"id":"l1.gdn.conv","from":28,"to":29,"format":"f32","rows":"n","dim":"lin_k_heads*lin_k_dim*2+lin_v_heads*lin_v_dim","fused":true},{"id":"l1.gdn.conv","from":28,"to":30,"format":"f32","rows":"n","dim":"lin_k_heads*lin_k_dim*2+lin_v_heads*lin_v_dim","fused":true},{"id":"l1.gdn.qkv","from":30,"to":31,"format":"f32","rows":"n","dim":"lin_k_heads*lin_k_dim*2+lin_v_heads*lin_v_dim","fused":false},{"id":"l1.gdn.core","from":31,"to":32,"format":"f32","rows":"n","dim":"lin_v_heads*lin_v_dim","fused":false},{"id":"l1.gdn.gated","from":32,"to":33,"format":"bf16","rows":"n","dim":"lin_v_heads*lin_v_dim","fused":false},{"id":"l1.gdn.o","from":33,"to":34,"format":"bf16","rows":"n","dim":"hidden","fused":false},{"id":"l1.gdn.y","from":34,"to":35,"format":"bf16","rows":"n","dim":"hidden","fused":false},{"id":"l1.gdn.y","from":34,"to":46,"format":"bf16","rows":"n","dim":"hidden","fused":false},{"id":"l1.moe_ffn.xn","from":35,"to":36,"format":"bf16","rows":"n","dim":"hidden","fused":false},{"id":"l1.moe_ffn.xn","from":35,"to":38,"format":"bf16","rows":"n","dim":"hidden","fused":false},{"id":"l1.moe_ffn.xn","from":35,"to":39,"format":"bf16","rows":"n","dim":"hidden","fused":false},{"id":"l1.moe_ffn.xn","from":35,"to":44,"format":"bf16","rows":"n","dim":"hidden","fused":false},{"id":"l1.moe_ffn.router_logits","from":36,"to":37,"format":"bf16","rows":"n","dim":"experts","fused":false},{"id":"l1.moe_ffn.topk_w","from":37,"to":45,"format":"f32","rows":"n","dim":"top_k","fused":false},{"id":"l1.moe_ffn.topk_id","from":37,"to":38,"format":"i32","rows":"n","dim":"top_k","fused":false},{"id":"l1.moe_ffn.topk_id","from":37,"to":41,"format":"i32","rows":"n","dim":"top_k","fused":false},{"id":"l1.moe_ffn.egu","from":38,"to":40,"format":"bf16","rows":"n*top_k","dim":"moe_inter*2","fused":false},{"id":"l1.moe_ffn.sgu","from":39,"to":42,"format":"bf16","rows":"n","dim":"shared_inter*2","fused":false},{"id":"l1.moe_ffn.eact","from":40,"to":41,"format":"f32","rows":"n*top_k","dim":"moe_inter","fused":true},{"id":"l1.moe_ffn.edown","from":41,"to":45,"format":"bf16","rows":"n*top_k","dim":"hidden","fused":false},{"id":"l1.moe_ffn.sact","from":42,"to":43,"format":"f32","rows":"n","dim":"shared_inter","fused":true},{"id":"l1.moe_ffn.sdown","from":43,"to":45,"format":"bf16","rows":"n","dim":"hidden","fused":false},{"id":"l1.moe_ffn.sg","from":44,"to":45,"format":"f32","rows":"n","dim":"1","fused":true},{"id":"l1.moe_ffn.f","from":45,"to":46,"format":"bf16","rows":"n","dim":"hidden","fused":false},{"id":"l1.moe_ffn.y","from":46,"to":47,"format":"bf16","rows":"n","dim":"hidden","fused":false},{"id":"l1.moe_ffn.y","from":46,"to":57,"format":"bf16","rows":"n","dim":"hidden","fused":false},{"id":"l2.gdn.xn","from":47,"to":48,"format":"bf16","rows":"n","dim":"hidden","fused":false},{"id":"l2.gdn.xn","from":47,"to":49,"format":"bf16","rows":"n","dim":"hidden","fused":false},{"id":"l2.gdn.qkvz","from":48,"to":51,"format":"bf16","rows":"n","dim":"lin_k_heads*lin_k_dim*2+lin_v_heads*lin_v_dim*2","fused":false},{"id":"l2.gdn.qkvz","from":48,"to":55,"format":"bf16","rows":"n","dim":"lin_k_heads*lin_k_dim*2+lin_v_heads*lin_v_dim*2","fused":false},{"id":"l2.gdn.ba","from":49,"to":50,"format":"f32","rows":"n","dim":"lin_v_heads*2","fused":true},{"id":"l2.gdn.decay","from":50,"to":54,"format":"f32","rows":"n","dim":"lin_v_heads","fused":false},{"id":"l2.gdn.beta","from":50,"to":54,"format":"f32","rows":"n","dim":"lin_v_heads","fused":false},{"id":"l2.gdn.conv","from":51,"to":52,"format":"f32","rows":"n","dim":"lin_k_heads*lin_k_dim*2+lin_v_heads*lin_v_dim","fused":true},{"id":"l2.gdn.conv","from":51,"to":53,"format":"f32","rows":"n","dim":"lin_k_heads*lin_k_dim*2+lin_v_heads*lin_v_dim","fused":true},{"id":"l2.gdn.qkv","from":53,"to":54,"format":"f32","rows":"n","dim":"lin_k_heads*lin_k_dim*2+lin_v_heads*lin_v_dim","fused":false},{"id":"l2.gdn.core","from":54,"to":55,"format":"f32","rows":"n","dim":"lin_v_heads*lin_v_dim","fused":false},{"id":"l2.gdn.gated","from":55,"to":56,"format":"bf16","rows":"n","dim":"lin_v_heads*lin_v_dim","fused":false},{"id":"l2.gdn.o","from":56,"to":57,"format":"bf16","rows":"n","dim":"hidden","fused":false},{"id":"l2.gdn.y","from":57,"to":58,"format":"bf16","rows":"n","dim":"hidden","fused":false},{"id":"l2.gdn.y","from":57,"to":69,"format":"bf16","rows":"n","dim":"hidden","fused":false},{"id":"l2.moe_ffn.xn","from":58,"to":59,"format":"bf16","rows":"n","dim":"hidden","fused":false},{"id":"l2.moe_ffn.xn","from":58,"to":61,"format":"bf16","rows":"n","dim":"hidden","fused":false},{"id":"l2.moe_ffn.xn","from":58,"to":62,"format":"bf16","rows":"n","dim":"hidden","fused":false},{"id":"l2.moe_ffn.xn","from":58,"to":67,"format":"bf16","rows":"n","dim":"hidden","fused":false},{"id":"l2.moe_ffn.router_logits","from":59,"to":60,"format":"bf16","rows":"n","dim":"experts","fused":false},{"id":"l2.moe_ffn.topk_w","from":60,"to":68,"format":"f32","rows":"n","dim":"top_k","fused":false},{"id":"l2.moe_ffn.topk_id","from":60,"to":61,"format":"i32","rows":"n","dim":"top_k","fused":false},{"id":"l2.moe_ffn.topk_id","from":60,"to":64,"format":"i32","rows":"n","dim":"top_k","fused":false},{"id":"l2.moe_ffn.egu","from":61,"to":63,"format":"bf16","rows":"n*top_k","dim":"moe_inter*2","fused":false},{"id":"l2.moe_ffn.sgu","from":62,"to":65,"format":"bf16","rows":"n","dim":"shared_inter*2","fused":false},{"id":"l2.moe_ffn.eact","from":63,"to":64,"format":"f32","rows":"n*top_k","dim":"moe_inter","fused":true},{"id":"l2.moe_ffn.edown","from":64,"to":68,"format":"bf16","rows":"n*top_k","dim":"hidden","fused":false},{"id":"l2.moe_ffn.sact","from":65,"to":66,"format":"f32","rows":"n","dim":"shared_inter","fused":true},{"id":"l2.moe_ffn.sdown","from":66,"to":68,"format":"bf16","rows":"n","dim":"hidden","fused":false},{"id":"l2.moe_ffn.sg","from":67,"to":68,"format":"f32","rows":"n","dim":"1","fused":true},{"id":"l2.moe_ffn.f","from":68,"to":69,"format":"bf16","rows":"n","dim":"hidden","fused":false},{"id":"l2.moe_ffn.y","from":69,"to":70,"format":"bf16","rows":"n","dim":"hidden","fused":false},{"id":"l2.moe_ffn.y","from":69,"to":82,"format":"bf16","rows":"n","dim":"hidden","fused":false},{"id":"l3.attn.xn","from":70,"to":71,"format":"bf16","rows":"n","dim":"hidden","fused":false},{"id":"l3.attn.xn","from":70,"to":73,"format":"bf16","rows":"n","dim":"hidden","fused":false},{"id":"l3.attn.xn","from":70,"to":74,"format":"bf16","rows":"n","dim":"hidden","fused":false},{"id":"l3.attn.qg","from":71,"to":72,"format":"bf16","rows":"n","dim":"q_heads*head_dim*2","fused":false},{"id":"l3.attn.q","from":72,"to":75,"format":"bf16","rows":"n","dim":"q_heads*head_dim","fused":false},{"id":"l3.attn.gate","from":72,"to":80,"format":"bf16","rows":"n","dim":"q_heads*head_dim","fused":false},{"id":"l3.attn.k","from":73,"to":76,"format":"bf16","rows":"n","dim":"kv_heads*head_dim","fused":false},{"id":"l3.attn.v","from":74,"to":78,"format":"bf16","rows":"n","dim":"kv_heads*head_dim","fused":false},{"id":"l3.attn.v","from":74,"to":79,"format":"bf16","rows":"n","dim":"kv_heads*head_dim","fused":false},{"id":"l3.attn.qn","from":75,"to":77,"format":"bf16","rows":"n","dim":"q_heads*head_dim","fused":false},{"id":"l3.attn.kn","from":76,"to":77,"format":"bf16","rows":"n","dim":"kv_heads*head_dim","fused":false},{"id":"l3.attn.qr","from":77,"to":79,"format":"bf16","rows":"n","dim":"q_heads*head_dim","fused":false},{"id":"l3.attn.kr","from":77,"to":78,"format":"bf16","rows":"n","dim":"kv_heads*head_dim","fused":false},{"id":"l3.attn.kr","from":77,"to":79,"format":"bf16","rows":"n","dim":"kv_heads*head_dim","fused":false},{"id":"l3.attn.a","from":79,"to":80,"format":"bf16","rows":"n","dim":"q_heads*head_dim","fused":false},{"id":"l3.attn.ag","from":80,"to":81,"format":"bf16","rows":"n","dim":"q_heads*head_dim","fused":false},{"id":"l3.attn.o","from":81,"to":82,"format":"bf16","rows":"n","dim":"hidden","fused":false},{"id":"l3.attn.y","from":82,"to":83,"format":"bf16","rows":"n","dim":"hidden","fused":false},{"id":"l3.attn.y","from":82,"to":94,"format":"bf16","rows":"n","dim":"hidden","fused":false},{"id":"l3.moe_ffn.xn","from":83,"to":84,"format":"bf16","rows":"n","dim":"hidden","fused":false},{"id":"l3.moe_ffn.xn","from":83,"to":86,"format":"bf16","rows":"n","dim":"hidden","fused":false},{"id":"l3.moe_ffn.xn","from":83,"to":87,"format":"bf16","rows":"n","dim":"hidden","fused":false},{"id":"l3.moe_ffn.xn","from":83,"to":92,"format":"bf16","rows":"n","dim":"hidden","fused":false},{"id":"l3.moe_ffn.router_logits","from":84,"to":85,"format":"bf16","rows":"n","dim":"experts","fused":false},{"id":"l3.moe_ffn.topk_w","from":85,"to":93,"format":"f32","rows":"n","dim":"top_k","fused":false},{"id":"l3.moe_ffn.topk_id","from":85,"to":86,"format":"i32","rows":"n","dim":"top_k","fused":false},{"id":"l3.moe_ffn.topk_id","from":85,"to":89,"format":"i32","rows":"n","dim":"top_k","fused":false},{"id":"l3.moe_ffn.egu","from":86,"to":88,"format":"bf16","rows":"n*top_k","dim":"moe_inter*2","fused":false},{"id":"l3.moe_ffn.sgu","from":87,"to":90,"format":"bf16","rows":"n","dim":"shared_inter*2","fused":false},{"id":"l3.moe_ffn.eact","from":88,"to":89,"format":"f32","rows":"n*top_k","dim":"moe_inter","fused":true},{"id":"l3.moe_ffn.edown","from":89,"to":93,"format":"bf16","rows":"n*top_k","dim":"hidden","fused":false},{"id":"l3.moe_ffn.sact","from":90,"to":91,"format":"f32","rows":"n","dim":"shared_inter","fused":true},{"id":"l3.moe_ffn.sdown","from":91,"to":93,"format":"bf16","rows":"n","dim":"hidden","fused":false},{"id":"l3.moe_ffn.sg","from":92,"to":93,"format":"f32","rows":"n","dim":"1","fused":true},{"id":"l3.moe_ffn.f","from":93,"to":94,"format":"bf16","rows":"n","dim":"hidden","fused":false},{"id":"l3.moe_ffn.y","from":94,"to":95,"format":"bf16","rows":"n","dim":"hidden","fused":false},{"id":"l3.moe_ffn.y","from":94,"to":105,"format":"bf16","rows":"n","dim":"hidden","fused":false},{"id":"l4.gdn.xn","from":95,"to":96,"format":"bf16","rows":"n","dim":"hidden","fused":false},{"id":"l4.gdn.xn","from":95,"to":97,"format":"bf16","rows":"n","dim":"hidden","fused":false},{"id":"l4.gdn.qkvz","from":96,"to":99,"format":"bf16","rows":"n","dim":"lin_k_heads*lin_k_dim*2+lin_v_heads*lin_v_dim*2","fused":false},{"id":"l4.gdn.qkvz","from":96,"to":103,"format":"bf16","rows":"n","dim":"lin_k_heads*lin_k_dim*2+lin_v_heads*lin_v_dim*2","fused":false},{"id":"l4.gdn.ba","from":97,"to":98,"format":"f32","rows":"n","dim":"lin_v_heads*2","fused":true},{"id":"l4.gdn.decay","from":98,"to":102,"format":"f32","rows":"n","dim":"lin_v_heads","fused":false},{"id":"l4.gdn.beta","from":98,"to":102,"format":"f32","rows":"n","dim":"lin_v_heads","fused":false},{"id":"l4.gdn.conv","from":99,"to":100,"format":"f32","rows":"n","dim":"lin_k_heads*lin_k_dim*2+lin_v_heads*lin_v_dim","fused":true},{"id":"l4.gdn.conv","from":99,"to":101,"format":"f32","rows":"n","dim":"lin_k_heads*lin_k_dim*2+lin_v_heads*lin_v_dim","fused":true},{"id":"l4.gdn.qkv","from":101,"to":102,"format":"f32","rows":"n","dim":"lin_k_heads*lin_k_dim*2+lin_v_heads*lin_v_dim","fused":false},{"id":"l4.gdn.core","from":102,"to":103,"format":"f32","rows":"n","dim":"lin_v_heads*lin_v_dim","fused":false},{"id":"l4.gdn.gated","from":103,"to":104,"format":"bf16","rows":"n","dim":"lin_v_heads*lin_v_dim","fused":false},{"id":"l4.gdn.o","from":104,"to":105,"format":"bf16","rows":"n","dim":"hidden","fused":false},{"id":"l4.gdn.y","from":105,"to":106,"format":"bf16","rows":"n","dim":"hidden","fused":false},{"id":"l4.gdn.y","from":105,"to":117,"format":"bf16","rows":"n","dim":"hidden","fused":false},{"id":"l4.moe_ffn.xn","from":106,"to":107,"format":"bf16","rows":"n","dim":"hidden","fused":false},{"id":"l4.moe_ffn.xn","from":106,"to":109,"format":"bf16","rows":"n","dim":"hidden","fused":false},{"id":"l4.moe_ffn.xn","from":106,"to":110,"format":"bf16","rows":"n","dim":"hidden","fused":false},{"id":"l4.moe_ffn.xn","from":106,"to":115,"format":"bf16","rows":"n","dim":"hidden","fused":false},{"id":"l4.moe_ffn.router_logits","from":107,"to":108,"format":"bf16","rows":"n","dim":"experts","fused":false},{"id":"l4.moe_ffn.topk_w","from":108,"to":116,"format":"f32","rows":"n","dim":"top_k","fused":false},{"id":"l4.moe_ffn.topk_id","from":108,"to":109,"format":"i32","rows":"n","dim":"top_k","fused":false},{"id":"l4.moe_ffn.topk_id","from":108,"to":112,"format":"i32","rows":"n","dim":"top_k","fused":false},{"id":"l4.moe_ffn.egu","from":109,"to":111,"format":"bf16","rows":"n*top_k","dim":"moe_inter*2","fused":false},{"id":"l4.moe_ffn.sgu","from":110,"to":113,"format":"bf16","rows":"n","dim":"shared_inter*2","fused":false},{"id":"l4.moe_ffn.eact","from":111,"to":112,"format":"f32","rows":"n*top_k","dim":"moe_inter","fused":true},{"id":"l4.moe_ffn.edown","from":112,"to":116,"format":"bf16","rows":"n*top_k","dim":"hidden","fused":false},{"id":"l4.moe_ffn.sact","from":113,"to":114,"format":"f32","rows":"n","dim":"shared_inter","fused":true},{"id":"l4.moe_ffn.sdown","from":114,"to":116,"format":"bf16","rows":"n","dim":"hidden","fused":false},{"id":"l4.moe_ffn.sg","from":115,"to":116,"format":"f32","rows":"n","dim":"1","fused":true},{"id":"l4.moe_ffn.f","from":116,"to":117,"format":"bf16","rows":"n","dim":"hidden","fused":false},{"id":"l4.moe_ffn.y","from":117,"to":118,"format":"bf16","rows":"n","dim":"hidden","fused":false},{"id":"l4.moe_ffn.y","from":117,"to":128,"format":"bf16","rows":"n","dim":"hidden","fused":false},{"id":"l5.gdn.xn","from":118,"to":119,"format":"bf16","rows":"n","dim":"hidden","fused":false},{"id":"l5.gdn.xn","from":118,"to":120,"format":"bf16","rows":"n","dim":"hidden","fused":false},{"id":"l5.gdn.qkvz","from":119,"to":122,"format":"bf16","rows":"n","dim":"lin_k_heads*lin_k_dim*2+lin_v_heads*lin_v_dim*2","fused":false},{"id":"l5.gdn.qkvz","from":119,"to":126,"format":"bf16","rows":"n","dim":"lin_k_heads*lin_k_dim*2+lin_v_heads*lin_v_dim*2","fused":false},{"id":"l5.gdn.ba","from":120,"to":121,"format":"f32","rows":"n","dim":"lin_v_heads*2","fused":true},{"id":"l5.gdn.decay","from":121,"to":125,"format":"f32","rows":"n","dim":"lin_v_heads","fused":false},{"id":"l5.gdn.beta","from":121,"to":125,"format":"f32","rows":"n","dim":"lin_v_heads","fused":false},{"id":"l5.gdn.conv","from":122,"to":123,"format":"f32","rows":"n","dim":"lin_k_heads*lin_k_dim*2+lin_v_heads*lin_v_dim","fused":true},{"id":"l5.gdn.conv","from":122,"to":124,"format":"f32","rows":"n","dim":"lin_k_heads*lin_k_dim*2+lin_v_heads*lin_v_dim","fused":true},{"id":"l5.gdn.qkv","from":124,"to":125,"format":"f32","rows":"n","dim":"lin_k_heads*lin_k_dim*2+lin_v_heads*lin_v_dim","fused":false},{"id":"l5.gdn.core","from":125,"to":126,"format":"f32","rows":"n","dim":"lin_v_heads*lin_v_dim","fused":false},{"id":"l5.gdn.gated","from":126,"to":127,"format":"bf16","rows":"n","dim":"lin_v_heads*lin_v_dim","fused":false},{"id":"l5.gdn.o","from":127,"to":128,"format":"bf16","rows":"n","dim":"hidden","fused":false},{"id":"l5.gdn.y","from":128,"to":129,"format":"bf16","rows":"n","dim":"hidden","fused":false},{"id":"l5.gdn.y","from":128,"to":140,"format":"bf16","rows":"n","dim":"hidden","fused":false},{"id":"l5.moe_ffn.xn","from":129,"to":130,"format":"bf16","rows":"n","dim":"hidden","fused":false},{"id":"l5.moe_ffn.xn","from":129,"to":132,"format":"bf16","rows":"n","dim":"hidden","fused":false},{"id":"l5.moe_ffn.xn","from":129,"to":133,"format":"bf16","rows":"n","dim":"hidden","fused":false},{"id":"l5.moe_ffn.xn","from":129,"to":138,"format":"bf16","rows":"n","dim":"hidden","fused":false},{"id":"l5.moe_ffn.router_logits","from":130,"to":131,"format":"bf16","rows":"n","dim":"experts","fused":false},{"id":"l5.moe_ffn.topk_w","from":131,"to":139,"format":"f32","rows":"n","dim":"top_k","fused":false},{"id":"l5.moe_ffn.topk_id","from":131,"to":132,"format":"i32","rows":"n","dim":"top_k","fused":false},{"id":"l5.moe_ffn.topk_id","from":131,"to":135,"format":"i32","rows":"n","dim":"top_k","fused":false},{"id":"l5.moe_ffn.egu","from":132,"to":134,"format":"bf16","rows":"n*top_k","dim":"moe_inter*2","fused":false},{"id":"l5.moe_ffn.sgu","from":133,"to":136,"format":"bf16","rows":"n","dim":"shared_inter*2","fused":false},{"id":"l5.moe_ffn.eact","from":134,"to":135,"format":"f32","rows":"n*top_k","dim":"moe_inter","fused":true},{"id":"l5.moe_ffn.edown","from":135,"to":139,"format":"bf16","rows":"n*top_k","dim":"hidden","fused":false},{"id":"l5.moe_ffn.sact","from":136,"to":137,"format":"f32","rows":"n","dim":"shared_inter","fused":true},{"id":"l5.moe_ffn.sdown","from":137,"to":139,"format":"bf16","rows":"n","dim":"hidden","fused":false},{"id":"l5.moe_ffn.sg","from":138,"to":139,"format":"f32","rows":"n","dim":"1","fused":true},{"id":"l5.moe_ffn.f","from":139,"to":140,"format":"bf16","rows":"n","dim":"hidden","fused":false},{"id":"l5.moe_ffn.y","from":140,"to":141,"format":"bf16","rows":"n","dim":"hidden","fused":false},{"id":"l5.moe_ffn.y","from":140,"to":151,"format":"bf16","rows":"n","dim":"hidden","fused":false},{"id":"l6.gdn.xn","from":141,"to":142,"format":"bf16","rows":"n","dim":"hidden","fused":false},{"id":"l6.gdn.xn","from":141,"to":143,"format":"bf16","rows":"n","dim":"hidden","fused":false},{"id":"l6.gdn.qkvz","from":142,"to":145,"format":"bf16","rows":"n","dim":"lin_k_heads*lin_k_dim*2+lin_v_heads*lin_v_dim*2","fused":false},{"id":"l6.gdn.qkvz","from":142,"to":149,"format":"bf16","rows":"n","dim":"lin_k_heads*lin_k_dim*2+lin_v_heads*lin_v_dim*2","fused":false},{"id":"l6.gdn.ba","from":143,"to":144,"format":"f32","rows":"n","dim":"lin_v_heads*2","fused":true},{"id":"l6.gdn.decay","from":144,"to":148,"format":"f32","rows":"n","dim":"lin_v_heads","fused":false},{"id":"l6.gdn.beta","from":144,"to":148,"format":"f32","rows":"n","dim":"lin_v_heads","fused":false},{"id":"l6.gdn.conv","from":145,"to":146,"format":"f32","rows":"n","dim":"lin_k_heads*lin_k_dim*2+lin_v_heads*lin_v_dim","fused":true},{"id":"l6.gdn.conv","from":145,"to":147,"format":"f32","rows":"n","dim":"lin_k_heads*lin_k_dim*2+lin_v_heads*lin_v_dim","fused":true},{"id":"l6.gdn.qkv","from":147,"to":148,"format":"f32","rows":"n","dim":"lin_k_heads*lin_k_dim*2+lin_v_heads*lin_v_dim","fused":false},{"id":"l6.gdn.core","from":148,"to":149,"format":"f32","rows":"n","dim":"lin_v_heads*lin_v_dim","fused":false},{"id":"l6.gdn.gated","from":149,"to":150,"format":"bf16","rows":"n","dim":"lin_v_heads*lin_v_dim","fused":false},{"id":"l6.gdn.o","from":150,"to":151,"format":"bf16","rows":"n","dim":"hidden","fused":false},{"id":"l6.gdn.y","from":151,"to":152,"format":"bf16","rows":"n","dim":"hidden","fused":false},{"id":"l6.gdn.y","from":151,"to":163,"format":"bf16","rows":"n","dim":"hidden","fused":false},{"id":"l6.moe_ffn.xn","from":152,"to":153,"format":"bf16","rows":"n","dim":"hidden","fused":false},{"id":"l6.moe_ffn.xn","from":152,"to":155,"format":"bf16","rows":"n","dim":"hidden","fused":false},{"id":"l6.moe_ffn.xn","from":152,"to":156,"format":"bf16","rows":"n","dim":"hidden","fused":false},{"id":"l6.moe_ffn.xn","from":152,"to":161,"format":"bf16","rows":"n","dim":"hidden","fused":false},{"id":"l6.moe_ffn.router_logits","from":153,"to":154,"format":"bf16","rows":"n","dim":"experts","fused":false},{"id":"l6.moe_ffn.topk_w","from":154,"to":162,"format":"f32","rows":"n","dim":"top_k","fused":false},{"id":"l6.moe_ffn.topk_id","from":154,"to":155,"format":"i32","rows":"n","dim":"top_k","fused":false},{"id":"l6.moe_ffn.topk_id","from":154,"to":158,"format":"i32","rows":"n","dim":"top_k","fused":false},{"id":"l6.moe_ffn.egu","from":155,"to":157,"format":"bf16","rows":"n*top_k","dim":"moe_inter*2","fused":false},{"id":"l6.moe_ffn.sgu","from":156,"to":159,"format":"bf16","rows":"n","dim":"shared_inter*2","fused":false},{"id":"l6.moe_ffn.eact","from":157,"to":158,"format":"f32","rows":"n*top_k","dim":"moe_inter","fused":true},{"id":"l6.moe_ffn.edown","from":158,"to":162,"format":"bf16","rows":"n*top_k","dim":"hidden","fused":false},{"id":"l6.moe_ffn.sact","from":159,"to":160,"format":"f32","rows":"n","dim":"shared_inter","fused":true},{"id":"l6.moe_ffn.sdown","from":160,"to":162,"format":"bf16","rows":"n","dim":"hidden","fused":false},{"id":"l6.moe_ffn.sg","from":161,"to":162,"format":"f32","rows":"n","dim":"1","fused":true},{"id":"l6.moe_ffn.f","from":162,"to":163,"format":"bf16","rows":"n","dim":"hidden","fused":false},{"id":"l6.moe_ffn.y","from":163,"to":164,"format":"bf16","rows":"n","dim":"hidden","fused":false},{"id":"l6.moe_ffn.y","from":163,"to":176,"format":"bf16","rows":"n","dim":"hidden","fused":false},{"id":"l7.attn.xn","from":164,"to":165,"format":"bf16","rows":"n","dim":"hidden","fused":false},{"id":"l7.attn.xn","from":164,"to":167,"format":"bf16","rows":"n","dim":"hidden","fused":false},{"id":"l7.attn.xn","from":164,"to":168,"format":"bf16","rows":"n","dim":"hidden","fused":false},{"id":"l7.attn.qg","from":165,"to":166,"format":"bf16","rows":"n","dim":"q_heads*head_dim*2","fused":false},{"id":"l7.attn.q","from":166,"to":169,"format":"bf16","rows":"n","dim":"q_heads*head_dim","fused":false},{"id":"l7.attn.gate","from":166,"to":174,"format":"bf16","rows":"n","dim":"q_heads*head_dim","fused":false},{"id":"l7.attn.k","from":167,"to":170,"format":"bf16","rows":"n","dim":"kv_heads*head_dim","fused":false},{"id":"l7.attn.v","from":168,"to":172,"format":"bf16","rows":"n","dim":"kv_heads*head_dim","fused":false},{"id":"l7.attn.v","from":168,"to":173,"format":"bf16","rows":"n","dim":"kv_heads*head_dim","fused":false},{"id":"l7.attn.qn","from":169,"to":171,"format":"bf16","rows":"n","dim":"q_heads*head_dim","fused":false},{"id":"l7.attn.kn","from":170,"to":171,"format":"bf16","rows":"n","dim":"kv_heads*head_dim","fused":false},{"id":"l7.attn.qr","from":171,"to":173,"format":"bf16","rows":"n","dim":"q_heads*head_dim","fused":false},{"id":"l7.attn.kr","from":171,"to":172,"format":"bf16","rows":"n","dim":"kv_heads*head_dim","fused":false},{"id":"l7.attn.kr","from":171,"to":173,"format":"bf16","rows":"n","dim":"kv_heads*head_dim","fused":false},{"id":"l7.attn.a","from":173,"to":174,"format":"bf16","rows":"n","dim":"q_heads*head_dim","fused":false},{"id":"l7.attn.ag","from":174,"to":175,"format":"bf16","rows":"n","dim":"q_heads*head_dim","fused":false},{"id":"l7.attn.o","from":175,"to":176,"format":"bf16","rows":"n","dim":"hidden","fused":false},{"id":"l7.attn.y","from":176,"to":177,"format":"bf16","rows":"n","dim":"hidden","fused":false},{"id":"l7.attn.y","from":176,"to":188,"format":"bf16","rows":"n","dim":"hidden","fused":false},{"id":"l7.moe_ffn.xn","from":177,"to":178,"format":"bf16","rows":"n","dim":"hidden","fused":false},{"id":"l7.moe_ffn.xn","from":177,"to":180,"format":"bf16","rows":"n","dim":"hidden","fused":false},{"id":"l7.moe_ffn.xn","from":177,"to":181,"format":"bf16","rows":"n","dim":"hidden","fused":false},{"id":"l7.moe_ffn.xn","from":177,"to":186,"format":"bf16","rows":"n","dim":"hidden","fused":false},{"id":"l7.moe_ffn.router_logits","from":178,"to":179,"format":"bf16","rows":"n","dim":"experts","fused":false},{"id":"l7.moe_ffn.topk_w","from":179,"to":187,"format":"f32","rows":"n","dim":"top_k","fused":false},{"id":"l7.moe_ffn.topk_id","from":179,"to":180,"format":"i32","rows":"n","dim":"top_k","fused":false},{"id":"l7.moe_ffn.topk_id","from":179,"to":183,"format":"i32","rows":"n","dim":"top_k","fused":false},{"id":"l7.moe_ffn.egu","from":180,"to":182,"format":"bf16","rows":"n*top_k","dim":"moe_inter*2","fused":false},{"id":"l7.moe_ffn.sgu","from":181,"to":184,"format":"bf16","rows":"n","dim":"shared_inter*2","fused":false},{"id":"l7.moe_ffn.eact","from":182,"to":183,"format":"f32","rows":"n*top_k","dim":"moe_inter","fused":true},{"id":"l7.moe_ffn.edown","from":183,"to":187,"format":"bf16","rows":"n*top_k","dim":"hidden","fused":false},{"id":"l7.moe_ffn.sact","from":184,"to":185,"format":"f32","rows":"n","dim":"shared_inter","fused":true},{"id":"l7.moe_ffn.sdown","from":185,"to":187,"format":"bf16","rows":"n","dim":"hidden","fused":false},{"id":"l7.moe_ffn.sg","from":186,"to":187,"format":"f32","rows":"n","dim":"1","fused":true},{"id":"l7.moe_ffn.f","from":187,"to":188,"format":"bf16","rows":"n","dim":"hidden","fused":false},{"id":"l7.moe_ffn.y","from":188,"to":189,"format":"bf16","rows":"n","dim":"hidden","fused":false},{"id":"l7.moe_ffn.y","from":188,"to":199,"format":"bf16","rows":"n","dim":"hidden","fused":false},{"id":"l8.gdn.xn","from":189,"to":190,"format":"bf16","rows":"n","dim":"hidden","fused":false},{"id":"l8.gdn.xn","from":189,"to":191,"format":"bf16","rows":"n","dim":"hidden","fused":false},{"id":"l8.gdn.qkvz","from":190,"to":193,"format":"bf16","rows":"n","dim":"lin_k_heads*lin_k_dim*2+lin_v_heads*lin_v_dim*2","fused":false},{"id":"l8.gdn.qkvz","from":190,"to":197,"format":"bf16","rows":"n","dim":"lin_k_heads*lin_k_dim*2+lin_v_heads*lin_v_dim*2","fused":false},{"id":"l8.gdn.ba","from":191,"to":192,"format":"f32","rows":"n","dim":"lin_v_heads*2","fused":true},{"id":"l8.gdn.decay","from":192,"to":196,"format":"f32","rows":"n","dim":"lin_v_heads","fused":false},{"id":"l8.gdn.beta","from":192,"to":196,"format":"f32","rows":"n","dim":"lin_v_heads","fused":false},{"id":"l8.gdn.conv","from":193,"to":194,"format":"f32","rows":"n","dim":"lin_k_heads*lin_k_dim*2+lin_v_heads*lin_v_dim","fused":true},{"id":"l8.gdn.conv","from":193,"to":195,"format":"f32","rows":"n","dim":"lin_k_heads*lin_k_dim*2+lin_v_heads*lin_v_dim","fused":true},{"id":"l8.gdn.qkv","from":195,"to":196,"format":"f32","rows":"n","dim":"lin_k_heads*lin_k_dim*2+lin_v_heads*lin_v_dim","fused":false},{"id":"l8.gdn.core","from":196,"to":197,"format":"f32","rows":"n","dim":"lin_v_heads*lin_v_dim","fused":false},{"id":"l8.gdn.gated","from":197,"to":198,"format":"bf16","rows":"n","dim":"lin_v_heads*lin_v_dim","fused":false},{"id":"l8.gdn.o","from":198,"to":199,"format":"bf16","rows":"n","dim":"hidden","fused":false},{"id":"l8.gdn.y","from":199,"to":200,"format":"bf16","rows":"n","dim":"hidden","fused":false},{"id":"l8.gdn.y","from":199,"to":211,"format":"bf16","rows":"n","dim":"hidden","fused":false},{"id":"l8.moe_ffn.xn","from":200,"to":201,"format":"bf16","rows":"n","dim":"hidden","fused":false},{"id":"l8.moe_ffn.xn","from":200,"to":203,"format":"bf16","rows":"n","dim":"hidden","fused":false},{"id":"l8.moe_ffn.xn","from":200,"to":204,"format":"bf16","rows":"n","dim":"hidden","fused":false},{"id":"l8.moe_ffn.xn","from":200,"to":209,"format":"bf16","rows":"n","dim":"hidden","fused":false},{"id":"l8.moe_ffn.router_logits","from":201,"to":202,"format":"bf16","rows":"n","dim":"experts","fused":false},{"id":"l8.moe_ffn.topk_w","from":202,"to":210,"format":"f32","rows":"n","dim":"top_k","fused":false},{"id":"l8.moe_ffn.topk_id","from":202,"to":203,"format":"i32","rows":"n","dim":"top_k","fused":false},{"id":"l8.moe_ffn.topk_id","from":202,"to":206,"format":"i32","rows":"n","dim":"top_k","fused":false},{"id":"l8.moe_ffn.egu","from":203,"to":205,"format":"bf16","rows":"n*top_k","dim":"moe_inter*2","fused":false},{"id":"l8.moe_ffn.sgu","from":204,"to":207,"format":"bf16","rows":"n","dim":"shared_inter*2","fused":false},{"id":"l8.moe_ffn.eact","from":205,"to":206,"format":"f32","rows":"n*top_k","dim":"moe_inter","fused":true},{"id":"l8.moe_ffn.edown","from":206,"to":210,"format":"bf16","rows":"n*top_k","dim":"hidden","fused":false},{"id":"l8.moe_ffn.sact","from":207,"to":208,"format":"f32","rows":"n","dim":"shared_inter","fused":true},{"id":"l8.moe_ffn.sdown","from":208,"to":210,"format":"bf16","rows":"n","dim":"hidden","fused":false},{"id":"l8.moe_ffn.sg","from":209,"to":210,"format":"f32","rows":"n","dim":"1","fused":true},{"id":"l8.moe_ffn.f","from":210,"to":211,"format":"bf16","rows":"n","dim":"hidden","fused":false},{"id":"l8.moe_ffn.y","from":211,"to":212,"format":"bf16","rows":"n","dim":"hidden","fused":false},{"id":"l8.moe_ffn.y","from":211,"to":222,"format":"bf16","rows":"n","dim":"hidden","fused":false},{"id":"l9.gdn.xn","from":212,"to":213,"format":"bf16","rows":"n","dim":"hidden","fused":false},{"id":"l9.gdn.xn","from":212,"to":214,"format":"bf16","rows":"n","dim":"hidden","fused":false},{"id":"l9.gdn.qkvz","from":213,"to":216,"format":"bf16","rows":"n","dim":"lin_k_heads*lin_k_dim*2+lin_v_heads*lin_v_dim*2","fused":false},{"id":"l9.gdn.qkvz","from":213,"to":220,"format":"bf16","rows":"n","dim":"lin_k_heads*lin_k_dim*2+lin_v_heads*lin_v_dim*2","fused":false},{"id":"l9.gdn.ba","from":214,"to":215,"format":"f32","rows":"n","dim":"lin_v_heads*2","fused":true},{"id":"l9.gdn.decay","from":215,"to":219,"format":"f32","rows":"n","dim":"lin_v_heads","fused":false},{"id":"l9.gdn.beta","from":215,"to":219,"format":"f32","rows":"n","dim":"lin_v_heads","fused":false},{"id":"l9.gdn.conv","from":216,"to":217,"format":"f32","rows":"n","dim":"lin_k_heads*lin_k_dim*2+lin_v_heads*lin_v_dim","fused":true},{"id":"l9.gdn.conv","from":216,"to":218,"format":"f32","rows":"n","dim":"lin_k_heads*lin_k_dim*2+lin_v_heads*lin_v_dim","fused":true},{"id":"l9.gdn.qkv","from":218,"to":219,"format":"f32","rows":"n","dim":"lin_k_heads*lin_k_dim*2+lin_v_heads*lin_v_dim","fused":false},{"id":"l9.gdn.core","from":219,"to":220,"format":"f32","rows":"n","dim":"lin_v_heads*lin_v_dim","fused":false},{"id":"l9.gdn.gated","from":220,"to":221,"format":"bf16","rows":"n","dim":"lin_v_heads*lin_v_dim","fused":false},{"id":"l9.gdn.o","from":221,"to":222,"format":"bf16","rows":"n","dim":"hidden","fused":false},{"id":"l9.gdn.y","from":222,"to":223,"format":"bf16","rows":"n","dim":"hidden","fused":false},{"id":"l9.gdn.y","from":222,"to":234,"format":"bf16","rows":"n","dim":"hidden","fused":false},{"id":"l9.moe_ffn.xn","from":223,"to":224,"format":"bf16","rows":"n","dim":"hidden","fused":false},{"id":"l9.moe_ffn.xn","from":223,"to":226,"format":"bf16","rows":"n","dim":"hidden","fused":false},{"id":"l9.moe_ffn.xn","from":223,"to":227,"format":"bf16","rows":"n","dim":"hidden","fused":false},{"id":"l9.moe_ffn.xn","from":223,"to":232,"format":"bf16","rows":"n","dim":"hidden","fused":false},{"id":"l9.moe_ffn.router_logits","from":224,"to":225,"format":"bf16","rows":"n","dim":"experts","fused":false},{"id":"l9.moe_ffn.topk_w","from":225,"to":233,"format":"f32","rows":"n","dim":"top_k","fused":false},{"id":"l9.moe_ffn.topk_id","from":225,"to":226,"format":"i32","rows":"n","dim":"top_k","fused":false},{"id":"l9.moe_ffn.topk_id","from":225,"to":229,"format":"i32","rows":"n","dim":"top_k","fused":false},{"id":"l9.moe_ffn.egu","from":226,"to":228,"format":"bf16","rows":"n*top_k","dim":"moe_inter*2","fused":false},{"id":"l9.moe_ffn.sgu","from":227,"to":230,"format":"bf16","rows":"n","dim":"shared_inter*2","fused":false},{"id":"l9.moe_ffn.eact","from":228,"to":229,"format":"f32","rows":"n*top_k","dim":"moe_inter","fused":true},{"id":"l9.moe_ffn.edown","from":229,"to":233,"format":"bf16","rows":"n*top_k","dim":"hidden","fused":false},{"id":"l9.moe_ffn.sact","from":230,"to":231,"format":"f32","rows":"n","dim":"shared_inter","fused":true},{"id":"l9.moe_ffn.sdown","from":231,"to":233,"format":"bf16","rows":"n","dim":"hidden","fused":false},{"id":"l9.moe_ffn.sg","from":232,"to":233,"format":"f32","rows":"n","dim":"1","fused":true},{"id":"l9.moe_ffn.f","from":233,"to":234,"format":"bf16","rows":"n","dim":"hidden","fused":false},{"id":"l9.moe_ffn.y","from":234,"to":235,"format":"bf16","rows":"n","dim":"hidden","fused":false},{"id":"l9.moe_ffn.y","from":234,"to":245,"format":"bf16","rows":"n","dim":"hidden","fused":false},{"id":"l10.gdn.xn","from":235,"to":236,"format":"bf16","rows":"n","dim":"hidden","fused":false},{"id":"l10.gdn.xn","from":235,"to":237,"format":"bf16","rows":"n","dim":"hidden","fused":false},{"id":"l10.gdn.qkvz","from":236,"to":239,"format":"bf16","rows":"n","dim":"lin_k_heads*lin_k_dim*2+lin_v_heads*lin_v_dim*2","fused":false},{"id":"l10.gdn.qkvz","from":236,"to":243,"format":"bf16","rows":"n","dim":"lin_k_heads*lin_k_dim*2+lin_v_heads*lin_v_dim*2","fused":false},{"id":"l10.gdn.ba","from":237,"to":238,"format":"f32","rows":"n","dim":"lin_v_heads*2","fused":true},{"id":"l10.gdn.decay","from":238,"to":242,"format":"f32","rows":"n","dim":"lin_v_heads","fused":false},{"id":"l10.gdn.beta","from":238,"to":242,"format":"f32","rows":"n","dim":"lin_v_heads","fused":false},{"id":"l10.gdn.conv","from":239,"to":240,"format":"f32","rows":"n","dim":"lin_k_heads*lin_k_dim*2+lin_v_heads*lin_v_dim","fused":true},{"id":"l10.gdn.conv","from":239,"to":241,"format":"f32","rows":"n","dim":"lin_k_heads*lin_k_dim*2+lin_v_heads*lin_v_dim","fused":true},{"id":"l10.gdn.qkv","from":241,"to":242,"format":"f32","rows":"n","dim":"lin_k_heads*lin_k_dim*2+lin_v_heads*lin_v_dim","fused":false},{"id":"l10.gdn.core","from":242,"to":243,"format":"f32","rows":"n","dim":"lin_v_heads*lin_v_dim","fused":false},{"id":"l10.gdn.gated","from":243,"to":244,"format":"bf16","rows":"n","dim":"lin_v_heads*lin_v_dim","fused":false},{"id":"l10.gdn.o","from":244,"to":245,"format":"bf16","rows":"n","dim":"hidden","fused":false},{"id":"l10.gdn.y","from":245,"to":246,"format":"bf16","rows":"n","dim":"hidden","fused":false},{"id":"l10.gdn.y","from":245,"to":257,"format":"bf16","rows":"n","dim":"hidden","fused":false},{"id":"l10.moe_ffn.xn","from":246,"to":247,"format":"bf16","rows":"n","dim":"hidden","fused":false},{"id":"l10.moe_ffn.xn","from":246,"to":249,"format":"bf16","rows":"n","dim":"hidden","fused":false},{"id":"l10.moe_ffn.xn","from":246,"to":250,"format":"bf16","rows":"n","dim":"hidden","fused":false},{"id":"l10.moe_ffn.xn","from":246,"to":255,"format":"bf16","rows":"n","dim":"hidden","fused":false},{"id":"l10.moe_ffn.router_logits","from":247,"to":248,"format":"bf16","rows":"n","dim":"experts","fused":false},{"id":"l10.moe_ffn.topk_w","from":248,"to":256,"format":"f32","rows":"n","dim":"top_k","fused":false},{"id":"l10.moe_ffn.topk_id","from":248,"to":249,"format":"i32","rows":"n","dim":"top_k","fused":false},{"id":"l10.moe_ffn.topk_id","from":248,"to":252,"format":"i32","rows":"n","dim":"top_k","fused":false},{"id":"l10.moe_ffn.egu","from":249,"to":251,"format":"bf16","rows":"n*top_k","dim":"moe_inter*2","fused":false},{"id":"l10.moe_ffn.sgu","from":250,"to":253,"format":"bf16","rows":"n","dim":"shared_inter*2","fused":false},{"id":"l10.moe_ffn.eact","from":251,"to":252,"format":"f32","rows":"n*top_k","dim":"moe_inter","fused":true},{"id":"l10.moe_ffn.edown","from":252,"to":256,"format":"bf16","rows":"n*top_k","dim":"hidden","fused":false},{"id":"l10.moe_ffn.sact","from":253,"to":254,"format":"f32","rows":"n","dim":"shared_inter","fused":true},{"id":"l10.moe_ffn.sdown","from":254,"to":256,"format":"bf16","rows":"n","dim":"hidden","fused":false},{"id":"l10.moe_ffn.sg","from":255,"to":256,"format":"f32","rows":"n","dim":"1","fused":true},{"id":"l10.moe_ffn.f","from":256,"to":257,"format":"bf16","rows":"n","dim":"hidden","fused":false},{"id":"l10.moe_ffn.y","from":257,"to":258,"format":"bf16","rows":"n","dim":"hidden","fused":false},{"id":"l10.moe_ffn.y","from":257,"to":270,"format":"bf16","rows":"n","dim":"hidden","fused":false},{"id":"l11.attn.xn","from":258,"to":259,"format":"bf16","rows":"n","dim":"hidden","fused":false},{"id":"l11.attn.xn","from":258,"to":261,"format":"bf16","rows":"n","dim":"hidden","fused":false},{"id":"l11.attn.xn","from":258,"to":262,"format":"bf16","rows":"n","dim":"hidden","fused":false},{"id":"l11.attn.qg","from":259,"to":260,"format":"bf16","rows":"n","dim":"q_heads*head_dim*2","fused":false},{"id":"l11.attn.q","from":260,"to":263,"format":"bf16","rows":"n","dim":"q_heads*head_dim","fused":false},{"id":"l11.attn.gate","from":260,"to":268,"format":"bf16","rows":"n","dim":"q_heads*head_dim","fused":false},{"id":"l11.attn.k","from":261,"to":264,"format":"bf16","rows":"n","dim":"kv_heads*head_dim","fused":false},{"id":"l11.attn.v","from":262,"to":266,"format":"bf16","rows":"n","dim":"kv_heads*head_dim","fused":false},{"id":"l11.attn.v","from":262,"to":267,"format":"bf16","rows":"n","dim":"kv_heads*head_dim","fused":false},{"id":"l11.attn.qn","from":263,"to":265,"format":"bf16","rows":"n","dim":"q_heads*head_dim","fused":false},{"id":"l11.attn.kn","from":264,"to":265,"format":"bf16","rows":"n","dim":"kv_heads*head_dim","fused":false},{"id":"l11.attn.qr","from":265,"to":267,"format":"bf16","rows":"n","dim":"q_heads*head_dim","fused":false},{"id":"l11.attn.kr","from":265,"to":266,"format":"bf16","rows":"n","dim":"kv_heads*head_dim","fused":false},{"id":"l11.attn.kr","from":265,"to":267,"format":"bf16","rows":"n","dim":"kv_heads*head_dim","fused":false},{"id":"l11.attn.a","from":267,"to":268,"format":"bf16","rows":"n","dim":"q_heads*head_dim","fused":false},{"id":"l11.attn.ag","from":268,"to":269,"format":"bf16","rows":"n","dim":"q_heads*head_dim","fused":false},{"id":"l11.attn.o","from":269,"to":270,"format":"bf16","rows":"n","dim":"hidden","fused":false},{"id":"l11.attn.y","from":270,"to":271,"format":"bf16","rows":"n","dim":"hidden","fused":false},{"id":"l11.attn.y","from":270,"to":282,"format":"bf16","rows":"n","dim":"hidden","fused":false},{"id":"l11.moe_ffn.xn","from":271,"to":272,"format":"bf16","rows":"n","dim":"hidden","fused":false},{"id":"l11.moe_ffn.xn","from":271,"to":274,"format":"bf16","rows":"n","dim":"hidden","fused":false},{"id":"l11.moe_ffn.xn","from":271,"to":275,"format":"bf16","rows":"n","dim":"hidden","fused":false},{"id":"l11.moe_ffn.xn","from":271,"to":280,"format":"bf16","rows":"n","dim":"hidden","fused":false},{"id":"l11.moe_ffn.router_logits","from":272,"to":273,"format":"bf16","rows":"n","dim":"experts","fused":false},{"id":"l11.moe_ffn.topk_w","from":273,"to":281,"format":"f32","rows":"n","dim":"top_k","fused":false},{"id":"l11.moe_ffn.topk_id","from":273,"to":274,"format":"i32","rows":"n","dim":"top_k","fused":false},{"id":"l11.moe_ffn.topk_id","from":273,"to":277,"format":"i32","rows":"n","dim":"top_k","fused":false},{"id":"l11.moe_ffn.egu","from":274,"to":276,"format":"bf16","rows":"n*top_k","dim":"moe_inter*2","fused":false},{"id":"l11.moe_ffn.sgu","from":275,"to":278,"format":"bf16","rows":"n","dim":"shared_inter*2","fused":false},{"id":"l11.moe_ffn.eact","from":276,"to":277,"format":"f32","rows":"n*top_k","dim":"moe_inter","fused":true},{"id":"l11.moe_ffn.edown","from":277,"to":281,"format":"bf16","rows":"n*top_k","dim":"hidden","fused":false},{"id":"l11.moe_ffn.sact","from":278,"to":279,"format":"f32","rows":"n","dim":"shared_inter","fused":true},{"id":"l11.moe_ffn.sdown","from":279,"to":281,"format":"bf16","rows":"n","dim":"hidden","fused":false},{"id":"l11.moe_ffn.sg","from":280,"to":281,"format":"f32","rows":"n","dim":"1","fused":true},{"id":"l11.moe_ffn.f","from":281,"to":282,"format":"bf16","rows":"n","dim":"hidden","fused":false},{"id":"l11.moe_ffn.y","from":282,"to":283,"format":"bf16","rows":"n","dim":"hidden","fused":false},{"id":"l11.moe_ffn.y","from":282,"to":293,"format":"bf16","rows":"n","dim":"hidden","fused":false},{"id":"l12.gdn.xn","from":283,"to":284,"format":"bf16","rows":"n","dim":"hidden","fused":false},{"id":"l12.gdn.xn","from":283,"to":285,"format":"bf16","rows":"n","dim":"hidden","fused":false},{"id":"l12.gdn.qkvz","from":284,"to":287,"format":"bf16","rows":"n","dim":"lin_k_heads*lin_k_dim*2+lin_v_heads*lin_v_dim*2","fused":false},{"id":"l12.gdn.qkvz","from":284,"to":291,"format":"bf16","rows":"n","dim":"lin_k_heads*lin_k_dim*2+lin_v_heads*lin_v_dim*2","fused":false},{"id":"l12.gdn.ba","from":285,"to":286,"format":"f32","rows":"n","dim":"lin_v_heads*2","fused":true},{"id":"l12.gdn.decay","from":286,"to":290,"format":"f32","rows":"n","dim":"lin_v_heads","fused":false},{"id":"l12.gdn.beta","from":286,"to":290,"format":"f32","rows":"n","dim":"lin_v_heads","fused":false},{"id":"l12.gdn.conv","from":287,"to":288,"format":"f32","rows":"n","dim":"lin_k_heads*lin_k_dim*2+lin_v_heads*lin_v_dim","fused":true},{"id":"l12.gdn.conv","from":287,"to":289,"format":"f32","rows":"n","dim":"lin_k_heads*lin_k_dim*2+lin_v_heads*lin_v_dim","fused":true},{"id":"l12.gdn.qkv","from":289,"to":290,"format":"f32","rows":"n","dim":"lin_k_heads*lin_k_dim*2+lin_v_heads*lin_v_dim","fused":false},{"id":"l12.gdn.core","from":290,"to":291,"format":"f32","rows":"n","dim":"lin_v_heads*lin_v_dim","fused":false},{"id":"l12.gdn.gated","from":291,"to":292,"format":"bf16","rows":"n","dim":"lin_v_heads*lin_v_dim","fused":false},{"id":"l12.gdn.o","from":292,"to":293,"format":"bf16","rows":"n","dim":"hidden","fused":false},{"id":"l12.gdn.y","from":293,"to":294,"format":"bf16","rows":"n","dim":"hidden","fused":false},{"id":"l12.gdn.y","from":293,"to":305,"format":"bf16","rows":"n","dim":"hidden","fused":false},{"id":"l12.moe_ffn.xn","from":294,"to":295,"format":"bf16","rows":"n","dim":"hidden","fused":false},{"id":"l12.moe_ffn.xn","from":294,"to":297,"format":"bf16","rows":"n","dim":"hidden","fused":false},{"id":"l12.moe_ffn.xn","from":294,"to":298,"format":"bf16","rows":"n","dim":"hidden","fused":false},{"id":"l12.moe_ffn.xn","from":294,"to":303,"format":"bf16","rows":"n","dim":"hidden","fused":false},{"id":"l12.moe_ffn.router_logits","from":295,"to":296,"format":"bf16","rows":"n","dim":"experts","fused":false},{"id":"l12.moe_ffn.topk_w","from":296,"to":304,"format":"f32","rows":"n","dim":"top_k","fused":false},{"id":"l12.moe_ffn.topk_id","from":296,"to":297,"format":"i32","rows":"n","dim":"top_k","fused":false},{"id":"l12.moe_ffn.topk_id","from":296,"to":300,"format":"i32","rows":"n","dim":"top_k","fused":false},{"id":"l12.moe_ffn.egu","from":297,"to":299,"format":"bf16","rows":"n*top_k","dim":"moe_inter*2","fused":false},{"id":"l12.moe_ffn.sgu","from":298,"to":301,"format":"bf16","rows":"n","dim":"shared_inter*2","fused":false},{"id":"l12.moe_ffn.eact","from":299,"to":300,"format":"f32","rows":"n*top_k","dim":"moe_inter","fused":true},{"id":"l12.moe_ffn.edown","from":300,"to":304,"format":"bf16","rows":"n*top_k","dim":"hidden","fused":false},{"id":"l12.moe_ffn.sact","from":301,"to":302,"format":"f32","rows":"n","dim":"shared_inter","fused":true},{"id":"l12.moe_ffn.sdown","from":302,"to":304,"format":"bf16","rows":"n","dim":"hidden","fused":false},{"id":"l12.moe_ffn.sg","from":303,"to":304,"format":"f32","rows":"n","dim":"1","fused":true},{"id":"l12.moe_ffn.f","from":304,"to":305,"format":"bf16","rows":"n","dim":"hidden","fused":false},{"id":"l12.moe_ffn.y","from":305,"to":306,"format":"bf16","rows":"n","dim":"hidden","fused":false},{"id":"l12.moe_ffn.y","from":305,"to":316,"format":"bf16","rows":"n","dim":"hidden","fused":false},{"id":"l13.gdn.xn","from":306,"to":307,"format":"bf16","rows":"n","dim":"hidden","fused":false},{"id":"l13.gdn.xn","from":306,"to":308,"format":"bf16","rows":"n","dim":"hidden","fused":false},{"id":"l13.gdn.qkvz","from":307,"to":310,"format":"bf16","rows":"n","dim":"lin_k_heads*lin_k_dim*2+lin_v_heads*lin_v_dim*2","fused":false},{"id":"l13.gdn.qkvz","from":307,"to":314,"format":"bf16","rows":"n","dim":"lin_k_heads*lin_k_dim*2+lin_v_heads*lin_v_dim*2","fused":false},{"id":"l13.gdn.ba","from":308,"to":309,"format":"f32","rows":"n","dim":"lin_v_heads*2","fused":true},{"id":"l13.gdn.decay","from":309,"to":313,"format":"f32","rows":"n","dim":"lin_v_heads","fused":false},{"id":"l13.gdn.beta","from":309,"to":313,"format":"f32","rows":"n","dim":"lin_v_heads","fused":false},{"id":"l13.gdn.conv","from":310,"to":311,"format":"f32","rows":"n","dim":"lin_k_heads*lin_k_dim*2+lin_v_heads*lin_v_dim","fused":true},{"id":"l13.gdn.conv","from":310,"to":312,"format":"f32","rows":"n","dim":"lin_k_heads*lin_k_dim*2+lin_v_heads*lin_v_dim","fused":true},{"id":"l13.gdn.qkv","from":312,"to":313,"format":"f32","rows":"n","dim":"lin_k_heads*lin_k_dim*2+lin_v_heads*lin_v_dim","fused":false},{"id":"l13.gdn.core","from":313,"to":314,"format":"f32","rows":"n","dim":"lin_v_heads*lin_v_dim","fused":false},{"id":"l13.gdn.gated","from":314,"to":315,"format":"bf16","rows":"n","dim":"lin_v_heads*lin_v_dim","fused":false},{"id":"l13.gdn.o","from":315,"to":316,"format":"bf16","rows":"n","dim":"hidden","fused":false},{"id":"l13.gdn.y","from":316,"to":317,"format":"bf16","rows":"n","dim":"hidden","fused":false},{"id":"l13.gdn.y","from":316,"to":328,"format":"bf16","rows":"n","dim":"hidden","fused":false},{"id":"l13.moe_ffn.xn","from":317,"to":318,"format":"bf16","rows":"n","dim":"hidden","fused":false},{"id":"l13.moe_ffn.xn","from":317,"to":320,"format":"bf16","rows":"n","dim":"hidden","fused":false},{"id":"l13.moe_ffn.xn","from":317,"to":321,"format":"bf16","rows":"n","dim":"hidden","fused":false},{"id":"l13.moe_ffn.xn","from":317,"to":326,"format":"bf16","rows":"n","dim":"hidden","fused":false},{"id":"l13.moe_ffn.router_logits","from":318,"to":319,"format":"bf16","rows":"n","dim":"experts","fused":false},{"id":"l13.moe_ffn.topk_w","from":319,"to":327,"format":"f32","rows":"n","dim":"top_k","fused":false},{"id":"l13.moe_ffn.topk_id","from":319,"to":320,"format":"i32","rows":"n","dim":"top_k","fused":false},{"id":"l13.moe_ffn.topk_id","from":319,"to":323,"format":"i32","rows":"n","dim":"top_k","fused":false},{"id":"l13.moe_ffn.egu","from":320,"to":322,"format":"bf16","rows":"n*top_k","dim":"moe_inter*2","fused":false},{"id":"l13.moe_ffn.sgu","from":321,"to":324,"format":"bf16","rows":"n","dim":"shared_inter*2","fused":false},{"id":"l13.moe_ffn.eact","from":322,"to":323,"format":"f32","rows":"n*top_k","dim":"moe_inter","fused":true},{"id":"l13.moe_ffn.edown","from":323,"to":327,"format":"bf16","rows":"n*top_k","dim":"hidden","fused":false},{"id":"l13.moe_ffn.sact","from":324,"to":325,"format":"f32","rows":"n","dim":"shared_inter","fused":true},{"id":"l13.moe_ffn.sdown","from":325,"to":327,"format":"bf16","rows":"n","dim":"hidden","fused":false},{"id":"l13.moe_ffn.sg","from":326,"to":327,"format":"f32","rows":"n","dim":"1","fused":true},{"id":"l13.moe_ffn.f","from":327,"to":328,"format":"bf16","rows":"n","dim":"hidden","fused":false},{"id":"l13.moe_ffn.y","from":328,"to":329,"format":"bf16","rows":"n","dim":"hidden","fused":false},{"id":"l13.moe_ffn.y","from":328,"to":339,"format":"bf16","rows":"n","dim":"hidden","fused":false},{"id":"l14.gdn.xn","from":329,"to":330,"format":"bf16","rows":"n","dim":"hidden","fused":false},{"id":"l14.gdn.xn","from":329,"to":331,"format":"bf16","rows":"n","dim":"hidden","fused":false},{"id":"l14.gdn.qkvz","from":330,"to":333,"format":"bf16","rows":"n","dim":"lin_k_heads*lin_k_dim*2+lin_v_heads*lin_v_dim*2","fused":false},{"id":"l14.gdn.qkvz","from":330,"to":337,"format":"bf16","rows":"n","dim":"lin_k_heads*lin_k_dim*2+lin_v_heads*lin_v_dim*2","fused":false},{"id":"l14.gdn.ba","from":331,"to":332,"format":"f32","rows":"n","dim":"lin_v_heads*2","fused":true},{"id":"l14.gdn.decay","from":332,"to":336,"format":"f32","rows":"n","dim":"lin_v_heads","fused":false},{"id":"l14.gdn.beta","from":332,"to":336,"format":"f32","rows":"n","dim":"lin_v_heads","fused":false},{"id":"l14.gdn.conv","from":333,"to":334,"format":"f32","rows":"n","dim":"lin_k_heads*lin_k_dim*2+lin_v_heads*lin_v_dim","fused":true},{"id":"l14.gdn.conv","from":333,"to":335,"format":"f32","rows":"n","dim":"lin_k_heads*lin_k_dim*2+lin_v_heads*lin_v_dim","fused":true},{"id":"l14.gdn.qkv","from":335,"to":336,"format":"f32","rows":"n","dim":"lin_k_heads*lin_k_dim*2+lin_v_heads*lin_v_dim","fused":false},{"id":"l14.gdn.core","from":336,"to":337,"format":"f32","rows":"n","dim":"lin_v_heads*lin_v_dim","fused":false},{"id":"l14.gdn.gated","from":337,"to":338,"format":"bf16","rows":"n","dim":"lin_v_heads*lin_v_dim","fused":false},{"id":"l14.gdn.o","from":338,"to":339,"format":"bf16","rows":"n","dim":"hidden","fused":false},{"id":"l14.gdn.y","from":339,"to":340,"format":"bf16","rows":"n","dim":"hidden","fused":false},{"id":"l14.gdn.y","from":339,"to":351,"format":"bf16","rows":"n","dim":"hidden","fused":false},{"id":"l14.moe_ffn.xn","from":340,"to":341,"format":"bf16","rows":"n","dim":"hidden","fused":false},{"id":"l14.moe_ffn.xn","from":340,"to":343,"format":"bf16","rows":"n","dim":"hidden","fused":false},{"id":"l14.moe_ffn.xn","from":340,"to":344,"format":"bf16","rows":"n","dim":"hidden","fused":false},{"id":"l14.moe_ffn.xn","from":340,"to":349,"format":"bf16","rows":"n","dim":"hidden","fused":false},{"id":"l14.moe_ffn.router_logits","from":341,"to":342,"format":"bf16","rows":"n","dim":"experts","fused":false},{"id":"l14.moe_ffn.topk_w","from":342,"to":350,"format":"f32","rows":"n","dim":"top_k","fused":false},{"id":"l14.moe_ffn.topk_id","from":342,"to":343,"format":"i32","rows":"n","dim":"top_k","fused":false},{"id":"l14.moe_ffn.topk_id","from":342,"to":346,"format":"i32","rows":"n","dim":"top_k","fused":false},{"id":"l14.moe_ffn.egu","from":343,"to":345,"format":"bf16","rows":"n*top_k","dim":"moe_inter*2","fused":false},{"id":"l14.moe_ffn.sgu","from":344,"to":347,"format":"bf16","rows":"n","dim":"shared_inter*2","fused":false},{"id":"l14.moe_ffn.eact","from":345,"to":346,"format":"f32","rows":"n*top_k","dim":"moe_inter","fused":true},{"id":"l14.moe_ffn.edown","from":346,"to":350,"format":"bf16","rows":"n*top_k","dim":"hidden","fused":false},{"id":"l14.moe_ffn.sact","from":347,"to":348,"format":"f32","rows":"n","dim":"shared_inter","fused":true},{"id":"l14.moe_ffn.sdown","from":348,"to":350,"format":"bf16","rows":"n","dim":"hidden","fused":false},{"id":"l14.moe_ffn.sg","from":349,"to":350,"format":"f32","rows":"n","dim":"1","fused":true},{"id":"l14.moe_ffn.f","from":350,"to":351,"format":"bf16","rows":"n","dim":"hidden","fused":false},{"id":"l14.moe_ffn.y","from":351,"to":352,"format":"bf16","rows":"n","dim":"hidden","fused":false},{"id":"l14.moe_ffn.y","from":351,"to":364,"format":"bf16","rows":"n","dim":"hidden","fused":false},{"id":"l15.attn.xn","from":352,"to":353,"format":"bf16","rows":"n","dim":"hidden","fused":false},{"id":"l15.attn.xn","from":352,"to":355,"format":"bf16","rows":"n","dim":"hidden","fused":false},{"id":"l15.attn.xn","from":352,"to":356,"format":"bf16","rows":"n","dim":"hidden","fused":false},{"id":"l15.attn.qg","from":353,"to":354,"format":"bf16","rows":"n","dim":"q_heads*head_dim*2","fused":false},{"id":"l15.attn.q","from":354,"to":357,"format":"bf16","rows":"n","dim":"q_heads*head_dim","fused":false},{"id":"l15.attn.gate","from":354,"to":362,"format":"bf16","rows":"n","dim":"q_heads*head_dim","fused":false},{"id":"l15.attn.k","from":355,"to":358,"format":"bf16","rows":"n","dim":"kv_heads*head_dim","fused":false},{"id":"l15.attn.v","from":356,"to":360,"format":"bf16","rows":"n","dim":"kv_heads*head_dim","fused":false},{"id":"l15.attn.v","from":356,"to":361,"format":"bf16","rows":"n","dim":"kv_heads*head_dim","fused":false},{"id":"l15.attn.qn","from":357,"to":359,"format":"bf16","rows":"n","dim":"q_heads*head_dim","fused":false},{"id":"l15.attn.kn","from":358,"to":359,"format":"bf16","rows":"n","dim":"kv_heads*head_dim","fused":false},{"id":"l15.attn.qr","from":359,"to":361,"format":"bf16","rows":"n","dim":"q_heads*head_dim","fused":false},{"id":"l15.attn.kr","from":359,"to":360,"format":"bf16","rows":"n","dim":"kv_heads*head_dim","fused":false},{"id":"l15.attn.kr","from":359,"to":361,"format":"bf16","rows":"n","dim":"kv_heads*head_dim","fused":false},{"id":"l15.attn.a","from":361,"to":362,"format":"bf16","rows":"n","dim":"q_heads*head_dim","fused":false},{"id":"l15.attn.ag","from":362,"to":363,"format":"bf16","rows":"n","dim":"q_heads*head_dim","fused":false},{"id":"l15.attn.o","from":363,"to":364,"format":"bf16","rows":"n","dim":"hidden","fused":false},{"id":"l15.attn.y","from":364,"to":365,"format":"bf16","rows":"n","dim":"hidden","fused":false},{"id":"l15.attn.y","from":364,"to":376,"format":"bf16","rows":"n","dim":"hidden","fused":false},{"id":"l15.moe_ffn.xn","from":365,"to":366,"format":"bf16","rows":"n","dim":"hidden","fused":false},{"id":"l15.moe_ffn.xn","from":365,"to":368,"format":"bf16","rows":"n","dim":"hidden","fused":false},{"id":"l15.moe_ffn.xn","from":365,"to":369,"format":"bf16","rows":"n","dim":"hidden","fused":false},{"id":"l15.moe_ffn.xn","from":365,"to":374,"format":"bf16","rows":"n","dim":"hidden","fused":false},{"id":"l15.moe_ffn.router_logits","from":366,"to":367,"format":"bf16","rows":"n","dim":"experts","fused":false},{"id":"l15.moe_ffn.topk_w","from":367,"to":375,"format":"f32","rows":"n","dim":"top_k","fused":false},{"id":"l15.moe_ffn.topk_id","from":367,"to":368,"format":"i32","rows":"n","dim":"top_k","fused":false},{"id":"l15.moe_ffn.topk_id","from":367,"to":371,"format":"i32","rows":"n","dim":"top_k","fused":false},{"id":"l15.moe_ffn.egu","from":368,"to":370,"format":"bf16","rows":"n*top_k","dim":"moe_inter*2","fused":false},{"id":"l15.moe_ffn.sgu","from":369,"to":372,"format":"bf16","rows":"n","dim":"shared_inter*2","fused":false},{"id":"l15.moe_ffn.eact","from":370,"to":371,"format":"f32","rows":"n*top_k","dim":"moe_inter","fused":true},{"id":"l15.moe_ffn.edown","from":371,"to":375,"format":"bf16","rows":"n*top_k","dim":"hidden","fused":false},{"id":"l15.moe_ffn.sact","from":372,"to":373,"format":"f32","rows":"n","dim":"shared_inter","fused":true},{"id":"l15.moe_ffn.sdown","from":373,"to":375,"format":"bf16","rows":"n","dim":"hidden","fused":false},{"id":"l15.moe_ffn.sg","from":374,"to":375,"format":"f32","rows":"n","dim":"1","fused":true},{"id":"l15.moe_ffn.f","from":375,"to":376,"format":"bf16","rows":"n","dim":"hidden","fused":false},{"id":"l15.moe_ffn.y","from":376,"to":377,"format":"bf16","rows":"n","dim":"hidden","fused":false},{"id":"l15.moe_ffn.y","from":376,"to":387,"format":"bf16","rows":"n","dim":"hidden","fused":false},{"id":"l16.gdn.xn","from":377,"to":378,"format":"bf16","rows":"n","dim":"hidden","fused":false},{"id":"l16.gdn.xn","from":377,"to":379,"format":"bf16","rows":"n","dim":"hidden","fused":false},{"id":"l16.gdn.qkvz","from":378,"to":381,"format":"bf16","rows":"n","dim":"lin_k_heads*lin_k_dim*2+lin_v_heads*lin_v_dim*2","fused":false},{"id":"l16.gdn.qkvz","from":378,"to":385,"format":"bf16","rows":"n","dim":"lin_k_heads*lin_k_dim*2+lin_v_heads*lin_v_dim*2","fused":false},{"id":"l16.gdn.ba","from":379,"to":380,"format":"f32","rows":"n","dim":"lin_v_heads*2","fused":true},{"id":"l16.gdn.decay","from":380,"to":384,"format":"f32","rows":"n","dim":"lin_v_heads","fused":false},{"id":"l16.gdn.beta","from":380,"to":384,"format":"f32","rows":"n","dim":"lin_v_heads","fused":false},{"id":"l16.gdn.conv","from":381,"to":382,"format":"f32","rows":"n","dim":"lin_k_heads*lin_k_dim*2+lin_v_heads*lin_v_dim","fused":true},{"id":"l16.gdn.conv","from":381,"to":383,"format":"f32","rows":"n","dim":"lin_k_heads*lin_k_dim*2+lin_v_heads*lin_v_dim","fused":true},{"id":"l16.gdn.qkv","from":383,"to":384,"format":"f32","rows":"n","dim":"lin_k_heads*lin_k_dim*2+lin_v_heads*lin_v_dim","fused":false},{"id":"l16.gdn.core","from":384,"to":385,"format":"f32","rows":"n","dim":"lin_v_heads*lin_v_dim","fused":false},{"id":"l16.gdn.gated","from":385,"to":386,"format":"bf16","rows":"n","dim":"lin_v_heads*lin_v_dim","fused":false},{"id":"l16.gdn.o","from":386,"to":387,"format":"bf16","rows":"n","dim":"hidden","fused":false},{"id":"l16.gdn.y","from":387,"to":388,"format":"bf16","rows":"n","dim":"hidden","fused":false},{"id":"l16.gdn.y","from":387,"to":399,"format":"bf16","rows":"n","dim":"hidden","fused":false},{"id":"l16.moe_ffn.xn","from":388,"to":389,"format":"bf16","rows":"n","dim":"hidden","fused":false},{"id":"l16.moe_ffn.xn","from":388,"to":391,"format":"bf16","rows":"n","dim":"hidden","fused":false},{"id":"l16.moe_ffn.xn","from":388,"to":392,"format":"bf16","rows":"n","dim":"hidden","fused":false},{"id":"l16.moe_ffn.xn","from":388,"to":397,"format":"bf16","rows":"n","dim":"hidden","fused":false},{"id":"l16.moe_ffn.router_logits","from":389,"to":390,"format":"bf16","rows":"n","dim":"experts","fused":false},{"id":"l16.moe_ffn.topk_w","from":390,"to":398,"format":"f32","rows":"n","dim":"top_k","fused":false},{"id":"l16.moe_ffn.topk_id","from":390,"to":391,"format":"i32","rows":"n","dim":"top_k","fused":false},{"id":"l16.moe_ffn.topk_id","from":390,"to":394,"format":"i32","rows":"n","dim":"top_k","fused":false},{"id":"l16.moe_ffn.egu","from":391,"to":393,"format":"bf16","rows":"n*top_k","dim":"moe_inter*2","fused":false},{"id":"l16.moe_ffn.sgu","from":392,"to":395,"format":"bf16","rows":"n","dim":"shared_inter*2","fused":false},{"id":"l16.moe_ffn.eact","from":393,"to":394,"format":"f32","rows":"n*top_k","dim":"moe_inter","fused":true},{"id":"l16.moe_ffn.edown","from":394,"to":398,"format":"bf16","rows":"n*top_k","dim":"hidden","fused":false},{"id":"l16.moe_ffn.sact","from":395,"to":396,"format":"f32","rows":"n","dim":"shared_inter","fused":true},{"id":"l16.moe_ffn.sdown","from":396,"to":398,"format":"bf16","rows":"n","dim":"hidden","fused":false},{"id":"l16.moe_ffn.sg","from":397,"to":398,"format":"f32","rows":"n","dim":"1","fused":true},{"id":"l16.moe_ffn.f","from":398,"to":399,"format":"bf16","rows":"n","dim":"hidden","fused":false},{"id":"l16.moe_ffn.y","from":399,"to":400,"format":"bf16","rows":"n","dim":"hidden","fused":false},{"id":"l16.moe_ffn.y","from":399,"to":410,"format":"bf16","rows":"n","dim":"hidden","fused":false},{"id":"l17.gdn.xn","from":400,"to":401,"format":"bf16","rows":"n","dim":"hidden","fused":false},{"id":"l17.gdn.xn","from":400,"to":402,"format":"bf16","rows":"n","dim":"hidden","fused":false},{"id":"l17.gdn.qkvz","from":401,"to":404,"format":"bf16","rows":"n","dim":"lin_k_heads*lin_k_dim*2+lin_v_heads*lin_v_dim*2","fused":false},{"id":"l17.gdn.qkvz","from":401,"to":408,"format":"bf16","rows":"n","dim":"lin_k_heads*lin_k_dim*2+lin_v_heads*lin_v_dim*2","fused":false},{"id":"l17.gdn.ba","from":402,"to":403,"format":"f32","rows":"n","dim":"lin_v_heads*2","fused":true},{"id":"l17.gdn.decay","from":403,"to":407,"format":"f32","rows":"n","dim":"lin_v_heads","fused":false},{"id":"l17.gdn.beta","from":403,"to":407,"format":"f32","rows":"n","dim":"lin_v_heads","fused":false},{"id":"l17.gdn.conv","from":404,"to":405,"format":"f32","rows":"n","dim":"lin_k_heads*lin_k_dim*2+lin_v_heads*lin_v_dim","fused":true},{"id":"l17.gdn.conv","from":404,"to":406,"format":"f32","rows":"n","dim":"lin_k_heads*lin_k_dim*2+lin_v_heads*lin_v_dim","fused":true},{"id":"l17.gdn.qkv","from":406,"to":407,"format":"f32","rows":"n","dim":"lin_k_heads*lin_k_dim*2+lin_v_heads*lin_v_dim","fused":false},{"id":"l17.gdn.core","from":407,"to":408,"format":"f32","rows":"n","dim":"lin_v_heads*lin_v_dim","fused":false},{"id":"l17.gdn.gated","from":408,"to":409,"format":"bf16","rows":"n","dim":"lin_v_heads*lin_v_dim","fused":false},{"id":"l17.gdn.o","from":409,"to":410,"format":"bf16","rows":"n","dim":"hidden","fused":false},{"id":"l17.gdn.y","from":410,"to":411,"format":"bf16","rows":"n","dim":"hidden","fused":false},{"id":"l17.gdn.y","from":410,"to":422,"format":"bf16","rows":"n","dim":"hidden","fused":false},{"id":"l17.moe_ffn.xn","from":411,"to":412,"format":"bf16","rows":"n","dim":"hidden","fused":false},{"id":"l17.moe_ffn.xn","from":411,"to":414,"format":"bf16","rows":"n","dim":"hidden","fused":false},{"id":"l17.moe_ffn.xn","from":411,"to":415,"format":"bf16","rows":"n","dim":"hidden","fused":false},{"id":"l17.moe_ffn.xn","from":411,"to":420,"format":"bf16","rows":"n","dim":"hidden","fused":false},{"id":"l17.moe_ffn.router_logits","from":412,"to":413,"format":"bf16","rows":"n","dim":"experts","fused":false},{"id":"l17.moe_ffn.topk_w","from":413,"to":421,"format":"f32","rows":"n","dim":"top_k","fused":false},{"id":"l17.moe_ffn.topk_id","from":413,"to":414,"format":"i32","rows":"n","dim":"top_k","fused":false},{"id":"l17.moe_ffn.topk_id","from":413,"to":417,"format":"i32","rows":"n","dim":"top_k","fused":false},{"id":"l17.moe_ffn.egu","from":414,"to":416,"format":"bf16","rows":"n*top_k","dim":"moe_inter*2","fused":false},{"id":"l17.moe_ffn.sgu","from":415,"to":418,"format":"bf16","rows":"n","dim":"shared_inter*2","fused":false},{"id":"l17.moe_ffn.eact","from":416,"to":417,"format":"f32","rows":"n*top_k","dim":"moe_inter","fused":true},{"id":"l17.moe_ffn.edown","from":417,"to":421,"format":"bf16","rows":"n*top_k","dim":"hidden","fused":false},{"id":"l17.moe_ffn.sact","from":418,"to":419,"format":"f32","rows":"n","dim":"shared_inter","fused":true},{"id":"l17.moe_ffn.sdown","from":419,"to":421,"format":"bf16","rows":"n","dim":"hidden","fused":false},{"id":"l17.moe_ffn.sg","from":420,"to":421,"format":"f32","rows":"n","dim":"1","fused":true},{"id":"l17.moe_ffn.f","from":421,"to":422,"format":"bf16","rows":"n","dim":"hidden","fused":false},{"id":"l17.moe_ffn.y","from":422,"to":423,"format":"bf16","rows":"n","dim":"hidden","fused":false},{"id":"l17.moe_ffn.y","from":422,"to":433,"format":"bf16","rows":"n","dim":"hidden","fused":false},{"id":"l18.gdn.xn","from":423,"to":424,"format":"bf16","rows":"n","dim":"hidden","fused":false},{"id":"l18.gdn.xn","from":423,"to":425,"format":"bf16","rows":"n","dim":"hidden","fused":false},{"id":"l18.gdn.qkvz","from":424,"to":427,"format":"bf16","rows":"n","dim":"lin_k_heads*lin_k_dim*2+lin_v_heads*lin_v_dim*2","fused":false},{"id":"l18.gdn.qkvz","from":424,"to":431,"format":"bf16","rows":"n","dim":"lin_k_heads*lin_k_dim*2+lin_v_heads*lin_v_dim*2","fused":false},{"id":"l18.gdn.ba","from":425,"to":426,"format":"f32","rows":"n","dim":"lin_v_heads*2","fused":true},{"id":"l18.gdn.decay","from":426,"to":430,"format":"f32","rows":"n","dim":"lin_v_heads","fused":false},{"id":"l18.gdn.beta","from":426,"to":430,"format":"f32","rows":"n","dim":"lin_v_heads","fused":false},{"id":"l18.gdn.conv","from":427,"to":428,"format":"f32","rows":"n","dim":"lin_k_heads*lin_k_dim*2+lin_v_heads*lin_v_dim","fused":true},{"id":"l18.gdn.conv","from":427,"to":429,"format":"f32","rows":"n","dim":"lin_k_heads*lin_k_dim*2+lin_v_heads*lin_v_dim","fused":true},{"id":"l18.gdn.qkv","from":429,"to":430,"format":"f32","rows":"n","dim":"lin_k_heads*lin_k_dim*2+lin_v_heads*lin_v_dim","fused":false},{"id":"l18.gdn.core","from":430,"to":431,"format":"f32","rows":"n","dim":"lin_v_heads*lin_v_dim","fused":false},{"id":"l18.gdn.gated","from":431,"to":432,"format":"bf16","rows":"n","dim":"lin_v_heads*lin_v_dim","fused":false},{"id":"l18.gdn.o","from":432,"to":433,"format":"bf16","rows":"n","dim":"hidden","fused":false},{"id":"l18.gdn.y","from":433,"to":434,"format":"bf16","rows":"n","dim":"hidden","fused":false},{"id":"l18.gdn.y","from":433,"to":445,"format":"bf16","rows":"n","dim":"hidden","fused":false},{"id":"l18.moe_ffn.xn","from":434,"to":435,"format":"bf16","rows":"n","dim":"hidden","fused":false},{"id":"l18.moe_ffn.xn","from":434,"to":437,"format":"bf16","rows":"n","dim":"hidden","fused":false},{"id":"l18.moe_ffn.xn","from":434,"to":438,"format":"bf16","rows":"n","dim":"hidden","fused":false},{"id":"l18.moe_ffn.xn","from":434,"to":443,"format":"bf16","rows":"n","dim":"hidden","fused":false},{"id":"l18.moe_ffn.router_logits","from":435,"to":436,"format":"bf16","rows":"n","dim":"experts","fused":false},{"id":"l18.moe_ffn.topk_w","from":436,"to":444,"format":"f32","rows":"n","dim":"top_k","fused":false},{"id":"l18.moe_ffn.topk_id","from":436,"to":437,"format":"i32","rows":"n","dim":"top_k","fused":false},{"id":"l18.moe_ffn.topk_id","from":436,"to":440,"format":"i32","rows":"n","dim":"top_k","fused":false},{"id":"l18.moe_ffn.egu","from":437,"to":439,"format":"bf16","rows":"n*top_k","dim":"moe_inter*2","fused":false},{"id":"l18.moe_ffn.sgu","from":438,"to":441,"format":"bf16","rows":"n","dim":"shared_inter*2","fused":false},{"id":"l18.moe_ffn.eact","from":439,"to":440,"format":"f32","rows":"n*top_k","dim":"moe_inter","fused":true},{"id":"l18.moe_ffn.edown","from":440,"to":444,"format":"bf16","rows":"n*top_k","dim":"hidden","fused":false},{"id":"l18.moe_ffn.sact","from":441,"to":442,"format":"f32","rows":"n","dim":"shared_inter","fused":true},{"id":"l18.moe_ffn.sdown","from":442,"to":444,"format":"bf16","rows":"n","dim":"hidden","fused":false},{"id":"l18.moe_ffn.sg","from":443,"to":444,"format":"f32","rows":"n","dim":"1","fused":true},{"id":"l18.moe_ffn.f","from":444,"to":445,"format":"bf16","rows":"n","dim":"hidden","fused":false},{"id":"l18.moe_ffn.y","from":445,"to":446,"format":"bf16","rows":"n","dim":"hidden","fused":false},{"id":"l18.moe_ffn.y","from":445,"to":458,"format":"bf16","rows":"n","dim":"hidden","fused":false},{"id":"l19.attn.xn","from":446,"to":447,"format":"bf16","rows":"n","dim":"hidden","fused":false},{"id":"l19.attn.xn","from":446,"to":449,"format":"bf16","rows":"n","dim":"hidden","fused":false},{"id":"l19.attn.xn","from":446,"to":450,"format":"bf16","rows":"n","dim":"hidden","fused":false},{"id":"l19.attn.qg","from":447,"to":448,"format":"bf16","rows":"n","dim":"q_heads*head_dim*2","fused":false},{"id":"l19.attn.q","from":448,"to":451,"format":"bf16","rows":"n","dim":"q_heads*head_dim","fused":false},{"id":"l19.attn.gate","from":448,"to":456,"format":"bf16","rows":"n","dim":"q_heads*head_dim","fused":false},{"id":"l19.attn.k","from":449,"to":452,"format":"bf16","rows":"n","dim":"kv_heads*head_dim","fused":false},{"id":"l19.attn.v","from":450,"to":454,"format":"bf16","rows":"n","dim":"kv_heads*head_dim","fused":false},{"id":"l19.attn.v","from":450,"to":455,"format":"bf16","rows":"n","dim":"kv_heads*head_dim","fused":false},{"id":"l19.attn.qn","from":451,"to":453,"format":"bf16","rows":"n","dim":"q_heads*head_dim","fused":false},{"id":"l19.attn.kn","from":452,"to":453,"format":"bf16","rows":"n","dim":"kv_heads*head_dim","fused":false},{"id":"l19.attn.qr","from":453,"to":455,"format":"bf16","rows":"n","dim":"q_heads*head_dim","fused":false},{"id":"l19.attn.kr","from":453,"to":454,"format":"bf16","rows":"n","dim":"kv_heads*head_dim","fused":false},{"id":"l19.attn.kr","from":453,"to":455,"format":"bf16","rows":"n","dim":"kv_heads*head_dim","fused":false},{"id":"l19.attn.a","from":455,"to":456,"format":"bf16","rows":"n","dim":"q_heads*head_dim","fused":false},{"id":"l19.attn.ag","from":456,"to":457,"format":"bf16","rows":"n","dim":"q_heads*head_dim","fused":false},{"id":"l19.attn.o","from":457,"to":458,"format":"bf16","rows":"n","dim":"hidden","fused":false},{"id":"l19.attn.y","from":458,"to":459,"format":"bf16","rows":"n","dim":"hidden","fused":false},{"id":"l19.attn.y","from":458,"to":470,"format":"bf16","rows":"n","dim":"hidden","fused":false},{"id":"l19.moe_ffn.xn","from":459,"to":460,"format":"bf16","rows":"n","dim":"hidden","fused":false},{"id":"l19.moe_ffn.xn","from":459,"to":462,"format":"bf16","rows":"n","dim":"hidden","fused":false},{"id":"l19.moe_ffn.xn","from":459,"to":463,"format":"bf16","rows":"n","dim":"hidden","fused":false},{"id":"l19.moe_ffn.xn","from":459,"to":468,"format":"bf16","rows":"n","dim":"hidden","fused":false},{"id":"l19.moe_ffn.router_logits","from":460,"to":461,"format":"bf16","rows":"n","dim":"experts","fused":false},{"id":"l19.moe_ffn.topk_w","from":461,"to":469,"format":"f32","rows":"n","dim":"top_k","fused":false},{"id":"l19.moe_ffn.topk_id","from":461,"to":462,"format":"i32","rows":"n","dim":"top_k","fused":false},{"id":"l19.moe_ffn.topk_id","from":461,"to":465,"format":"i32","rows":"n","dim":"top_k","fused":false},{"id":"l19.moe_ffn.egu","from":462,"to":464,"format":"bf16","rows":"n*top_k","dim":"moe_inter*2","fused":false},{"id":"l19.moe_ffn.sgu","from":463,"to":466,"format":"bf16","rows":"n","dim":"shared_inter*2","fused":false},{"id":"l19.moe_ffn.eact","from":464,"to":465,"format":"f32","rows":"n*top_k","dim":"moe_inter","fused":true},{"id":"l19.moe_ffn.edown","from":465,"to":469,"format":"bf16","rows":"n*top_k","dim":"hidden","fused":false},{"id":"l19.moe_ffn.sact","from":466,"to":467,"format":"f32","rows":"n","dim":"shared_inter","fused":true},{"id":"l19.moe_ffn.sdown","from":467,"to":469,"format":"bf16","rows":"n","dim":"hidden","fused":false},{"id":"l19.moe_ffn.sg","from":468,"to":469,"format":"f32","rows":"n","dim":"1","fused":true},{"id":"l19.moe_ffn.f","from":469,"to":470,"format":"bf16","rows":"n","dim":"hidden","fused":false},{"id":"l19.moe_ffn.y","from":470,"to":471,"format":"bf16","rows":"n","dim":"hidden","fused":false},{"id":"l19.moe_ffn.y","from":470,"to":481,"format":"bf16","rows":"n","dim":"hidden","fused":false},{"id":"l20.gdn.xn","from":471,"to":472,"format":"bf16","rows":"n","dim":"hidden","fused":false},{"id":"l20.gdn.xn","from":471,"to":473,"format":"bf16","rows":"n","dim":"hidden","fused":false},{"id":"l20.gdn.qkvz","from":472,"to":475,"format":"bf16","rows":"n","dim":"lin_k_heads*lin_k_dim*2+lin_v_heads*lin_v_dim*2","fused":false},{"id":"l20.gdn.qkvz","from":472,"to":479,"format":"bf16","rows":"n","dim":"lin_k_heads*lin_k_dim*2+lin_v_heads*lin_v_dim*2","fused":false},{"id":"l20.gdn.ba","from":473,"to":474,"format":"f32","rows":"n","dim":"lin_v_heads*2","fused":true},{"id":"l20.gdn.decay","from":474,"to":478,"format":"f32","rows":"n","dim":"lin_v_heads","fused":false},{"id":"l20.gdn.beta","from":474,"to":478,"format":"f32","rows":"n","dim":"lin_v_heads","fused":false},{"id":"l20.gdn.conv","from":475,"to":476,"format":"f32","rows":"n","dim":"lin_k_heads*lin_k_dim*2+lin_v_heads*lin_v_dim","fused":true},{"id":"l20.gdn.conv","from":475,"to":477,"format":"f32","rows":"n","dim":"lin_k_heads*lin_k_dim*2+lin_v_heads*lin_v_dim","fused":true},{"id":"l20.gdn.qkv","from":477,"to":478,"format":"f32","rows":"n","dim":"lin_k_heads*lin_k_dim*2+lin_v_heads*lin_v_dim","fused":false},{"id":"l20.gdn.core","from":478,"to":479,"format":"f32","rows":"n","dim":"lin_v_heads*lin_v_dim","fused":false},{"id":"l20.gdn.gated","from":479,"to":480,"format":"bf16","rows":"n","dim":"lin_v_heads*lin_v_dim","fused":false},{"id":"l20.gdn.o","from":480,"to":481,"format":"bf16","rows":"n","dim":"hidden","fused":false},{"id":"l20.gdn.y","from":481,"to":482,"format":"bf16","rows":"n","dim":"hidden","fused":false},{"id":"l20.gdn.y","from":481,"to":493,"format":"bf16","rows":"n","dim":"hidden","fused":false},{"id":"l20.moe_ffn.xn","from":482,"to":483,"format":"bf16","rows":"n","dim":"hidden","fused":false},{"id":"l20.moe_ffn.xn","from":482,"to":485,"format":"bf16","rows":"n","dim":"hidden","fused":false},{"id":"l20.moe_ffn.xn","from":482,"to":486,"format":"bf16","rows":"n","dim":"hidden","fused":false},{"id":"l20.moe_ffn.xn","from":482,"to":491,"format":"bf16","rows":"n","dim":"hidden","fused":false},{"id":"l20.moe_ffn.router_logits","from":483,"to":484,"format":"bf16","rows":"n","dim":"experts","fused":false},{"id":"l20.moe_ffn.topk_w","from":484,"to":492,"format":"f32","rows":"n","dim":"top_k","fused":false},{"id":"l20.moe_ffn.topk_id","from":484,"to":485,"format":"i32","rows":"n","dim":"top_k","fused":false},{"id":"l20.moe_ffn.topk_id","from":484,"to":488,"format":"i32","rows":"n","dim":"top_k","fused":false},{"id":"l20.moe_ffn.egu","from":485,"to":487,"format":"bf16","rows":"n*top_k","dim":"moe_inter*2","fused":false},{"id":"l20.moe_ffn.sgu","from":486,"to":489,"format":"bf16","rows":"n","dim":"shared_inter*2","fused":false},{"id":"l20.moe_ffn.eact","from":487,"to":488,"format":"f32","rows":"n*top_k","dim":"moe_inter","fused":true},{"id":"l20.moe_ffn.edown","from":488,"to":492,"format":"bf16","rows":"n*top_k","dim":"hidden","fused":false},{"id":"l20.moe_ffn.sact","from":489,"to":490,"format":"f32","rows":"n","dim":"shared_inter","fused":true},{"id":"l20.moe_ffn.sdown","from":490,"to":492,"format":"bf16","rows":"n","dim":"hidden","fused":false},{"id":"l20.moe_ffn.sg","from":491,"to":492,"format":"f32","rows":"n","dim":"1","fused":true},{"id":"l20.moe_ffn.f","from":492,"to":493,"format":"bf16","rows":"n","dim":"hidden","fused":false},{"id":"l20.moe_ffn.y","from":493,"to":494,"format":"bf16","rows":"n","dim":"hidden","fused":false},{"id":"l20.moe_ffn.y","from":493,"to":504,"format":"bf16","rows":"n","dim":"hidden","fused":false},{"id":"l21.gdn.xn","from":494,"to":495,"format":"bf16","rows":"n","dim":"hidden","fused":false},{"id":"l21.gdn.xn","from":494,"to":496,"format":"bf16","rows":"n","dim":"hidden","fused":false},{"id":"l21.gdn.qkvz","from":495,"to":498,"format":"bf16","rows":"n","dim":"lin_k_heads*lin_k_dim*2+lin_v_heads*lin_v_dim*2","fused":false},{"id":"l21.gdn.qkvz","from":495,"to":502,"format":"bf16","rows":"n","dim":"lin_k_heads*lin_k_dim*2+lin_v_heads*lin_v_dim*2","fused":false},{"id":"l21.gdn.ba","from":496,"to":497,"format":"f32","rows":"n","dim":"lin_v_heads*2","fused":true},{"id":"l21.gdn.decay","from":497,"to":501,"format":"f32","rows":"n","dim":"lin_v_heads","fused":false},{"id":"l21.gdn.beta","from":497,"to":501,"format":"f32","rows":"n","dim":"lin_v_heads","fused":false},{"id":"l21.gdn.conv","from":498,"to":499,"format":"f32","rows":"n","dim":"lin_k_heads*lin_k_dim*2+lin_v_heads*lin_v_dim","fused":true},{"id":"l21.gdn.conv","from":498,"to":500,"format":"f32","rows":"n","dim":"lin_k_heads*lin_k_dim*2+lin_v_heads*lin_v_dim","fused":true},{"id":"l21.gdn.qkv","from":500,"to":501,"format":"f32","rows":"n","dim":"lin_k_heads*lin_k_dim*2+lin_v_heads*lin_v_dim","fused":false},{"id":"l21.gdn.core","from":501,"to":502,"format":"f32","rows":"n","dim":"lin_v_heads*lin_v_dim","fused":false},{"id":"l21.gdn.gated","from":502,"to":503,"format":"bf16","rows":"n","dim":"lin_v_heads*lin_v_dim","fused":false},{"id":"l21.gdn.o","from":503,"to":504,"format":"bf16","rows":"n","dim":"hidden","fused":false},{"id":"l21.gdn.y","from":504,"to":505,"format":"bf16","rows":"n","dim":"hidden","fused":false},{"id":"l21.gdn.y","from":504,"to":516,"format":"bf16","rows":"n","dim":"hidden","fused":false},{"id":"l21.moe_ffn.xn","from":505,"to":506,"format":"bf16","rows":"n","dim":"hidden","fused":false},{"id":"l21.moe_ffn.xn","from":505,"to":508,"format":"bf16","rows":"n","dim":"hidden","fused":false},{"id":"l21.moe_ffn.xn","from":505,"to":509,"format":"bf16","rows":"n","dim":"hidden","fused":false},{"id":"l21.moe_ffn.xn","from":505,"to":514,"format":"bf16","rows":"n","dim":"hidden","fused":false},{"id":"l21.moe_ffn.router_logits","from":506,"to":507,"format":"bf16","rows":"n","dim":"experts","fused":false},{"id":"l21.moe_ffn.topk_w","from":507,"to":515,"format":"f32","rows":"n","dim":"top_k","fused":false},{"id":"l21.moe_ffn.topk_id","from":507,"to":508,"format":"i32","rows":"n","dim":"top_k","fused":false},{"id":"l21.moe_ffn.topk_id","from":507,"to":511,"format":"i32","rows":"n","dim":"top_k","fused":false},{"id":"l21.moe_ffn.egu","from":508,"to":510,"format":"bf16","rows":"n*top_k","dim":"moe_inter*2","fused":false},{"id":"l21.moe_ffn.sgu","from":509,"to":512,"format":"bf16","rows":"n","dim":"shared_inter*2","fused":false},{"id":"l21.moe_ffn.eact","from":510,"to":511,"format":"f32","rows":"n*top_k","dim":"moe_inter","fused":true},{"id":"l21.moe_ffn.edown","from":511,"to":515,"format":"bf16","rows":"n*top_k","dim":"hidden","fused":false},{"id":"l21.moe_ffn.sact","from":512,"to":513,"format":"f32","rows":"n","dim":"shared_inter","fused":true},{"id":"l21.moe_ffn.sdown","from":513,"to":515,"format":"bf16","rows":"n","dim":"hidden","fused":false},{"id":"l21.moe_ffn.sg","from":514,"to":515,"format":"f32","rows":"n","dim":"1","fused":true},{"id":"l21.moe_ffn.f","from":515,"to":516,"format":"bf16","rows":"n","dim":"hidden","fused":false},{"id":"l21.moe_ffn.y","from":516,"to":517,"format":"bf16","rows":"n","dim":"hidden","fused":false},{"id":"l21.moe_ffn.y","from":516,"to":527,"format":"bf16","rows":"n","dim":"hidden","fused":false},{"id":"l22.gdn.xn","from":517,"to":518,"format":"bf16","rows":"n","dim":"hidden","fused":false},{"id":"l22.gdn.xn","from":517,"to":519,"format":"bf16","rows":"n","dim":"hidden","fused":false},{"id":"l22.gdn.qkvz","from":518,"to":521,"format":"bf16","rows":"n","dim":"lin_k_heads*lin_k_dim*2+lin_v_heads*lin_v_dim*2","fused":false},{"id":"l22.gdn.qkvz","from":518,"to":525,"format":"bf16","rows":"n","dim":"lin_k_heads*lin_k_dim*2+lin_v_heads*lin_v_dim*2","fused":false},{"id":"l22.gdn.ba","from":519,"to":520,"format":"f32","rows":"n","dim":"lin_v_heads*2","fused":true},{"id":"l22.gdn.decay","from":520,"to":524,"format":"f32","rows":"n","dim":"lin_v_heads","fused":false},{"id":"l22.gdn.beta","from":520,"to":524,"format":"f32","rows":"n","dim":"lin_v_heads","fused":false},{"id":"l22.gdn.conv","from":521,"to":522,"format":"f32","rows":"n","dim":"lin_k_heads*lin_k_dim*2+lin_v_heads*lin_v_dim","fused":true},{"id":"l22.gdn.conv","from":521,"to":523,"format":"f32","rows":"n","dim":"lin_k_heads*lin_k_dim*2+lin_v_heads*lin_v_dim","fused":true},{"id":"l22.gdn.qkv","from":523,"to":524,"format":"f32","rows":"n","dim":"lin_k_heads*lin_k_dim*2+lin_v_heads*lin_v_dim","fused":false},{"id":"l22.gdn.core","from":524,"to":525,"format":"f32","rows":"n","dim":"lin_v_heads*lin_v_dim","fused":false},{"id":"l22.gdn.gated","from":525,"to":526,"format":"bf16","rows":"n","dim":"lin_v_heads*lin_v_dim","fused":false},{"id":"l22.gdn.o","from":526,"to":527,"format":"bf16","rows":"n","dim":"hidden","fused":false},{"id":"l22.gdn.y","from":527,"to":528,"format":"bf16","rows":"n","dim":"hidden","fused":false},{"id":"l22.gdn.y","from":527,"to":539,"format":"bf16","rows":"n","dim":"hidden","fused":false},{"id":"l22.moe_ffn.xn","from":528,"to":529,"format":"bf16","rows":"n","dim":"hidden","fused":false},{"id":"l22.moe_ffn.xn","from":528,"to":531,"format":"bf16","rows":"n","dim":"hidden","fused":false},{"id":"l22.moe_ffn.xn","from":528,"to":532,"format":"bf16","rows":"n","dim":"hidden","fused":false},{"id":"l22.moe_ffn.xn","from":528,"to":537,"format":"bf16","rows":"n","dim":"hidden","fused":false},{"id":"l22.moe_ffn.router_logits","from":529,"to":530,"format":"bf16","rows":"n","dim":"experts","fused":false},{"id":"l22.moe_ffn.topk_w","from":530,"to":538,"format":"f32","rows":"n","dim":"top_k","fused":false},{"id":"l22.moe_ffn.topk_id","from":530,"to":531,"format":"i32","rows":"n","dim":"top_k","fused":false},{"id":"l22.moe_ffn.topk_id","from":530,"to":534,"format":"i32","rows":"n","dim":"top_k","fused":false},{"id":"l22.moe_ffn.egu","from":531,"to":533,"format":"bf16","rows":"n*top_k","dim":"moe_inter*2","fused":false},{"id":"l22.moe_ffn.sgu","from":532,"to":535,"format":"bf16","rows":"n","dim":"shared_inter*2","fused":false},{"id":"l22.moe_ffn.eact","from":533,"to":534,"format":"f32","rows":"n*top_k","dim":"moe_inter","fused":true},{"id":"l22.moe_ffn.edown","from":534,"to":538,"format":"bf16","rows":"n*top_k","dim":"hidden","fused":false},{"id":"l22.moe_ffn.sact","from":535,"to":536,"format":"f32","rows":"n","dim":"shared_inter","fused":true},{"id":"l22.moe_ffn.sdown","from":536,"to":538,"format":"bf16","rows":"n","dim":"hidden","fused":false},{"id":"l22.moe_ffn.sg","from":537,"to":538,"format":"f32","rows":"n","dim":"1","fused":true},{"id":"l22.moe_ffn.f","from":538,"to":539,"format":"bf16","rows":"n","dim":"hidden","fused":false},{"id":"l22.moe_ffn.y","from":539,"to":540,"format":"bf16","rows":"n","dim":"hidden","fused":false},{"id":"l22.moe_ffn.y","from":539,"to":552,"format":"bf16","rows":"n","dim":"hidden","fused":false},{"id":"l23.attn.xn","from":540,"to":541,"format":"bf16","rows":"n","dim":"hidden","fused":false},{"id":"l23.attn.xn","from":540,"to":543,"format":"bf16","rows":"n","dim":"hidden","fused":false},{"id":"l23.attn.xn","from":540,"to":544,"format":"bf16","rows":"n","dim":"hidden","fused":false},{"id":"l23.attn.qg","from":541,"to":542,"format":"bf16","rows":"n","dim":"q_heads*head_dim*2","fused":false},{"id":"l23.attn.q","from":542,"to":545,"format":"bf16","rows":"n","dim":"q_heads*head_dim","fused":false},{"id":"l23.attn.gate","from":542,"to":550,"format":"bf16","rows":"n","dim":"q_heads*head_dim","fused":false},{"id":"l23.attn.k","from":543,"to":546,"format":"bf16","rows":"n","dim":"kv_heads*head_dim","fused":false},{"id":"l23.attn.v","from":544,"to":548,"format":"bf16","rows":"n","dim":"kv_heads*head_dim","fused":false},{"id":"l23.attn.v","from":544,"to":549,"format":"bf16","rows":"n","dim":"kv_heads*head_dim","fused":false},{"id":"l23.attn.qn","from":545,"to":547,"format":"bf16","rows":"n","dim":"q_heads*head_dim","fused":false},{"id":"l23.attn.kn","from":546,"to":547,"format":"bf16","rows":"n","dim":"kv_heads*head_dim","fused":false},{"id":"l23.attn.qr","from":547,"to":549,"format":"bf16","rows":"n","dim":"q_heads*head_dim","fused":false},{"id":"l23.attn.kr","from":547,"to":548,"format":"bf16","rows":"n","dim":"kv_heads*head_dim","fused":false},{"id":"l23.attn.kr","from":547,"to":549,"format":"bf16","rows":"n","dim":"kv_heads*head_dim","fused":false},{"id":"l23.attn.a","from":549,"to":550,"format":"bf16","rows":"n","dim":"q_heads*head_dim","fused":false},{"id":"l23.attn.ag","from":550,"to":551,"format":"bf16","rows":"n","dim":"q_heads*head_dim","fused":false},{"id":"l23.attn.o","from":551,"to":552,"format":"bf16","rows":"n","dim":"hidden","fused":false},{"id":"l23.attn.y","from":552,"to":553,"format":"bf16","rows":"n","dim":"hidden","fused":false},{"id":"l23.attn.y","from":552,"to":564,"format":"bf16","rows":"n","dim":"hidden","fused":false},{"id":"l23.moe_ffn.xn","from":553,"to":554,"format":"bf16","rows":"n","dim":"hidden","fused":false},{"id":"l23.moe_ffn.xn","from":553,"to":556,"format":"bf16","rows":"n","dim":"hidden","fused":false},{"id":"l23.moe_ffn.xn","from":553,"to":557,"format":"bf16","rows":"n","dim":"hidden","fused":false},{"id":"l23.moe_ffn.xn","from":553,"to":562,"format":"bf16","rows":"n","dim":"hidden","fused":false},{"id":"l23.moe_ffn.router_logits","from":554,"to":555,"format":"bf16","rows":"n","dim":"experts","fused":false},{"id":"l23.moe_ffn.topk_w","from":555,"to":563,"format":"f32","rows":"n","dim":"top_k","fused":false},{"id":"l23.moe_ffn.topk_id","from":555,"to":556,"format":"i32","rows":"n","dim":"top_k","fused":false},{"id":"l23.moe_ffn.topk_id","from":555,"to":559,"format":"i32","rows":"n","dim":"top_k","fused":false},{"id":"l23.moe_ffn.egu","from":556,"to":558,"format":"bf16","rows":"n*top_k","dim":"moe_inter*2","fused":false},{"id":"l23.moe_ffn.sgu","from":557,"to":560,"format":"bf16","rows":"n","dim":"shared_inter*2","fused":false},{"id":"l23.moe_ffn.eact","from":558,"to":559,"format":"f32","rows":"n*top_k","dim":"moe_inter","fused":true},{"id":"l23.moe_ffn.edown","from":559,"to":563,"format":"bf16","rows":"n*top_k","dim":"hidden","fused":false},{"id":"l23.moe_ffn.sact","from":560,"to":561,"format":"f32","rows":"n","dim":"shared_inter","fused":true},{"id":"l23.moe_ffn.sdown","from":561,"to":563,"format":"bf16","rows":"n","dim":"hidden","fused":false},{"id":"l23.moe_ffn.sg","from":562,"to":563,"format":"f32","rows":"n","dim":"1","fused":true},{"id":"l23.moe_ffn.f","from":563,"to":564,"format":"bf16","rows":"n","dim":"hidden","fused":false},{"id":"l23.moe_ffn.y","from":564,"to":565,"format":"bf16","rows":"n","dim":"hidden","fused":false},{"id":"l23.moe_ffn.y","from":564,"to":575,"format":"bf16","rows":"n","dim":"hidden","fused":false},{"id":"l24.gdn.xn","from":565,"to":566,"format":"bf16","rows":"n","dim":"hidden","fused":false},{"id":"l24.gdn.xn","from":565,"to":567,"format":"bf16","rows":"n","dim":"hidden","fused":false},{"id":"l24.gdn.qkvz","from":566,"to":569,"format":"bf16","rows":"n","dim":"lin_k_heads*lin_k_dim*2+lin_v_heads*lin_v_dim*2","fused":false},{"id":"l24.gdn.qkvz","from":566,"to":573,"format":"bf16","rows":"n","dim":"lin_k_heads*lin_k_dim*2+lin_v_heads*lin_v_dim*2","fused":false},{"id":"l24.gdn.ba","from":567,"to":568,"format":"f32","rows":"n","dim":"lin_v_heads*2","fused":true},{"id":"l24.gdn.decay","from":568,"to":572,"format":"f32","rows":"n","dim":"lin_v_heads","fused":false},{"id":"l24.gdn.beta","from":568,"to":572,"format":"f32","rows":"n","dim":"lin_v_heads","fused":false},{"id":"l24.gdn.conv","from":569,"to":570,"format":"f32","rows":"n","dim":"lin_k_heads*lin_k_dim*2+lin_v_heads*lin_v_dim","fused":true},{"id":"l24.gdn.conv","from":569,"to":571,"format":"f32","rows":"n","dim":"lin_k_heads*lin_k_dim*2+lin_v_heads*lin_v_dim","fused":true},{"id":"l24.gdn.qkv","from":571,"to":572,"format":"f32","rows":"n","dim":"lin_k_heads*lin_k_dim*2+lin_v_heads*lin_v_dim","fused":false},{"id":"l24.gdn.core","from":572,"to":573,"format":"f32","rows":"n","dim":"lin_v_heads*lin_v_dim","fused":false},{"id":"l24.gdn.gated","from":573,"to":574,"format":"bf16","rows":"n","dim":"lin_v_heads*lin_v_dim","fused":false},{"id":"l24.gdn.o","from":574,"to":575,"format":"bf16","rows":"n","dim":"hidden","fused":false},{"id":"l24.gdn.y","from":575,"to":576,"format":"bf16","rows":"n","dim":"hidden","fused":false},{"id":"l24.gdn.y","from":575,"to":587,"format":"bf16","rows":"n","dim":"hidden","fused":false},{"id":"l24.moe_ffn.xn","from":576,"to":577,"format":"bf16","rows":"n","dim":"hidden","fused":false},{"id":"l24.moe_ffn.xn","from":576,"to":579,"format":"bf16","rows":"n","dim":"hidden","fused":false},{"id":"l24.moe_ffn.xn","from":576,"to":580,"format":"bf16","rows":"n","dim":"hidden","fused":false},{"id":"l24.moe_ffn.xn","from":576,"to":585,"format":"bf16","rows":"n","dim":"hidden","fused":false},{"id":"l24.moe_ffn.router_logits","from":577,"to":578,"format":"bf16","rows":"n","dim":"experts","fused":false},{"id":"l24.moe_ffn.topk_w","from":578,"to":586,"format":"f32","rows":"n","dim":"top_k","fused":false},{"id":"l24.moe_ffn.topk_id","from":578,"to":579,"format":"i32","rows":"n","dim":"top_k","fused":false},{"id":"l24.moe_ffn.topk_id","from":578,"to":582,"format":"i32","rows":"n","dim":"top_k","fused":false},{"id":"l24.moe_ffn.egu","from":579,"to":581,"format":"bf16","rows":"n*top_k","dim":"moe_inter*2","fused":false},{"id":"l24.moe_ffn.sgu","from":580,"to":583,"format":"bf16","rows":"n","dim":"shared_inter*2","fused":false},{"id":"l24.moe_ffn.eact","from":581,"to":582,"format":"f32","rows":"n*top_k","dim":"moe_inter","fused":true},{"id":"l24.moe_ffn.edown","from":582,"to":586,"format":"bf16","rows":"n*top_k","dim":"hidden","fused":false},{"id":"l24.moe_ffn.sact","from":583,"to":584,"format":"f32","rows":"n","dim":"shared_inter","fused":true},{"id":"l24.moe_ffn.sdown","from":584,"to":586,"format":"bf16","rows":"n","dim":"hidden","fused":false},{"id":"l24.moe_ffn.sg","from":585,"to":586,"format":"f32","rows":"n","dim":"1","fused":true},{"id":"l24.moe_ffn.f","from":586,"to":587,"format":"bf16","rows":"n","dim":"hidden","fused":false},{"id":"l24.moe_ffn.y","from":587,"to":588,"format":"bf16","rows":"n","dim":"hidden","fused":false},{"id":"l24.moe_ffn.y","from":587,"to":598,"format":"bf16","rows":"n","dim":"hidden","fused":false},{"id":"l25.gdn.xn","from":588,"to":589,"format":"bf16","rows":"n","dim":"hidden","fused":false},{"id":"l25.gdn.xn","from":588,"to":590,"format":"bf16","rows":"n","dim":"hidden","fused":false},{"id":"l25.gdn.qkvz","from":589,"to":592,"format":"bf16","rows":"n","dim":"lin_k_heads*lin_k_dim*2+lin_v_heads*lin_v_dim*2","fused":false},{"id":"l25.gdn.qkvz","from":589,"to":596,"format":"bf16","rows":"n","dim":"lin_k_heads*lin_k_dim*2+lin_v_heads*lin_v_dim*2","fused":false},{"id":"l25.gdn.ba","from":590,"to":591,"format":"f32","rows":"n","dim":"lin_v_heads*2","fused":true},{"id":"l25.gdn.decay","from":591,"to":595,"format":"f32","rows":"n","dim":"lin_v_heads","fused":false},{"id":"l25.gdn.beta","from":591,"to":595,"format":"f32","rows":"n","dim":"lin_v_heads","fused":false},{"id":"l25.gdn.conv","from":592,"to":593,"format":"f32","rows":"n","dim":"lin_k_heads*lin_k_dim*2+lin_v_heads*lin_v_dim","fused":true},{"id":"l25.gdn.conv","from":592,"to":594,"format":"f32","rows":"n","dim":"lin_k_heads*lin_k_dim*2+lin_v_heads*lin_v_dim","fused":true},{"id":"l25.gdn.qkv","from":594,"to":595,"format":"f32","rows":"n","dim":"lin_k_heads*lin_k_dim*2+lin_v_heads*lin_v_dim","fused":false},{"id":"l25.gdn.core","from":595,"to":596,"format":"f32","rows":"n","dim":"lin_v_heads*lin_v_dim","fused":false},{"id":"l25.gdn.gated","from":596,"to":597,"format":"bf16","rows":"n","dim":"lin_v_heads*lin_v_dim","fused":false},{"id":"l25.gdn.o","from":597,"to":598,"format":"bf16","rows":"n","dim":"hidden","fused":false},{"id":"l25.gdn.y","from":598,"to":599,"format":"bf16","rows":"n","dim":"hidden","fused":false},{"id":"l25.gdn.y","from":598,"to":610,"format":"bf16","rows":"n","dim":"hidden","fused":false},{"id":"l25.moe_ffn.xn","from":599,"to":600,"format":"bf16","rows":"n","dim":"hidden","fused":false},{"id":"l25.moe_ffn.xn","from":599,"to":602,"format":"bf16","rows":"n","dim":"hidden","fused":false},{"id":"l25.moe_ffn.xn","from":599,"to":603,"format":"bf16","rows":"n","dim":"hidden","fused":false},{"id":"l25.moe_ffn.xn","from":599,"to":608,"format":"bf16","rows":"n","dim":"hidden","fused":false},{"id":"l25.moe_ffn.router_logits","from":600,"to":601,"format":"bf16","rows":"n","dim":"experts","fused":false},{"id":"l25.moe_ffn.topk_w","from":601,"to":609,"format":"f32","rows":"n","dim":"top_k","fused":false},{"id":"l25.moe_ffn.topk_id","from":601,"to":602,"format":"i32","rows":"n","dim":"top_k","fused":false},{"id":"l25.moe_ffn.topk_id","from":601,"to":605,"format":"i32","rows":"n","dim":"top_k","fused":false},{"id":"l25.moe_ffn.egu","from":602,"to":604,"format":"bf16","rows":"n*top_k","dim":"moe_inter*2","fused":false},{"id":"l25.moe_ffn.sgu","from":603,"to":606,"format":"bf16","rows":"n","dim":"shared_inter*2","fused":false},{"id":"l25.moe_ffn.eact","from":604,"to":605,"format":"f32","rows":"n*top_k","dim":"moe_inter","fused":true},{"id":"l25.moe_ffn.edown","from":605,"to":609,"format":"bf16","rows":"n*top_k","dim":"hidden","fused":false},{"id":"l25.moe_ffn.sact","from":606,"to":607,"format":"f32","rows":"n","dim":"shared_inter","fused":true},{"id":"l25.moe_ffn.sdown","from":607,"to":609,"format":"bf16","rows":"n","dim":"hidden","fused":false},{"id":"l25.moe_ffn.sg","from":608,"to":609,"format":"f32","rows":"n","dim":"1","fused":true},{"id":"l25.moe_ffn.f","from":609,"to":610,"format":"bf16","rows":"n","dim":"hidden","fused":false},{"id":"l25.moe_ffn.y","from":610,"to":611,"format":"bf16","rows":"n","dim":"hidden","fused":false},{"id":"l25.moe_ffn.y","from":610,"to":621,"format":"bf16","rows":"n","dim":"hidden","fused":false},{"id":"l26.gdn.xn","from":611,"to":612,"format":"bf16","rows":"n","dim":"hidden","fused":false},{"id":"l26.gdn.xn","from":611,"to":613,"format":"bf16","rows":"n","dim":"hidden","fused":false},{"id":"l26.gdn.qkvz","from":612,"to":615,"format":"bf16","rows":"n","dim":"lin_k_heads*lin_k_dim*2+lin_v_heads*lin_v_dim*2","fused":false},{"id":"l26.gdn.qkvz","from":612,"to":619,"format":"bf16","rows":"n","dim":"lin_k_heads*lin_k_dim*2+lin_v_heads*lin_v_dim*2","fused":false},{"id":"l26.gdn.ba","from":613,"to":614,"format":"f32","rows":"n","dim":"lin_v_heads*2","fused":true},{"id":"l26.gdn.decay","from":614,"to":618,"format":"f32","rows":"n","dim":"lin_v_heads","fused":false},{"id":"l26.gdn.beta","from":614,"to":618,"format":"f32","rows":"n","dim":"lin_v_heads","fused":false},{"id":"l26.gdn.conv","from":615,"to":616,"format":"f32","rows":"n","dim":"lin_k_heads*lin_k_dim*2+lin_v_heads*lin_v_dim","fused":true},{"id":"l26.gdn.conv","from":615,"to":617,"format":"f32","rows":"n","dim":"lin_k_heads*lin_k_dim*2+lin_v_heads*lin_v_dim","fused":true},{"id":"l26.gdn.qkv","from":617,"to":618,"format":"f32","rows":"n","dim":"lin_k_heads*lin_k_dim*2+lin_v_heads*lin_v_dim","fused":false},{"id":"l26.gdn.core","from":618,"to":619,"format":"f32","rows":"n","dim":"lin_v_heads*lin_v_dim","fused":false},{"id":"l26.gdn.gated","from":619,"to":620,"format":"bf16","rows":"n","dim":"lin_v_heads*lin_v_dim","fused":false},{"id":"l26.gdn.o","from":620,"to":621,"format":"bf16","rows":"n","dim":"hidden","fused":false},{"id":"l26.gdn.y","from":621,"to":622,"format":"bf16","rows":"n","dim":"hidden","fused":false},{"id":"l26.gdn.y","from":621,"to":633,"format":"bf16","rows":"n","dim":"hidden","fused":false},{"id":"l26.moe_ffn.xn","from":622,"to":623,"format":"bf16","rows":"n","dim":"hidden","fused":false},{"id":"l26.moe_ffn.xn","from":622,"to":625,"format":"bf16","rows":"n","dim":"hidden","fused":false},{"id":"l26.moe_ffn.xn","from":622,"to":626,"format":"bf16","rows":"n","dim":"hidden","fused":false},{"id":"l26.moe_ffn.xn","from":622,"to":631,"format":"bf16","rows":"n","dim":"hidden","fused":false},{"id":"l26.moe_ffn.router_logits","from":623,"to":624,"format":"bf16","rows":"n","dim":"experts","fused":false},{"id":"l26.moe_ffn.topk_w","from":624,"to":632,"format":"f32","rows":"n","dim":"top_k","fused":false},{"id":"l26.moe_ffn.topk_id","from":624,"to":625,"format":"i32","rows":"n","dim":"top_k","fused":false},{"id":"l26.moe_ffn.topk_id","from":624,"to":628,"format":"i32","rows":"n","dim":"top_k","fused":false},{"id":"l26.moe_ffn.egu","from":625,"to":627,"format":"bf16","rows":"n*top_k","dim":"moe_inter*2","fused":false},{"id":"l26.moe_ffn.sgu","from":626,"to":629,"format":"bf16","rows":"n","dim":"shared_inter*2","fused":false},{"id":"l26.moe_ffn.eact","from":627,"to":628,"format":"f32","rows":"n*top_k","dim":"moe_inter","fused":true},{"id":"l26.moe_ffn.edown","from":628,"to":632,"format":"bf16","rows":"n*top_k","dim":"hidden","fused":false},{"id":"l26.moe_ffn.sact","from":629,"to":630,"format":"f32","rows":"n","dim":"shared_inter","fused":true},{"id":"l26.moe_ffn.sdown","from":630,"to":632,"format":"bf16","rows":"n","dim":"hidden","fused":false},{"id":"l26.moe_ffn.sg","from":631,"to":632,"format":"f32","rows":"n","dim":"1","fused":true},{"id":"l26.moe_ffn.f","from":632,"to":633,"format":"bf16","rows":"n","dim":"hidden","fused":false},{"id":"l26.moe_ffn.y","from":633,"to":634,"format":"bf16","rows":"n","dim":"hidden","fused":false},{"id":"l26.moe_ffn.y","from":633,"to":646,"format":"bf16","rows":"n","dim":"hidden","fused":false},{"id":"l27.attn.xn","from":634,"to":635,"format":"bf16","rows":"n","dim":"hidden","fused":false},{"id":"l27.attn.xn","from":634,"to":637,"format":"bf16","rows":"n","dim":"hidden","fused":false},{"id":"l27.attn.xn","from":634,"to":638,"format":"bf16","rows":"n","dim":"hidden","fused":false},{"id":"l27.attn.qg","from":635,"to":636,"format":"bf16","rows":"n","dim":"q_heads*head_dim*2","fused":false},{"id":"l27.attn.q","from":636,"to":639,"format":"bf16","rows":"n","dim":"q_heads*head_dim","fused":false},{"id":"l27.attn.gate","from":636,"to":644,"format":"bf16","rows":"n","dim":"q_heads*head_dim","fused":false},{"id":"l27.attn.k","from":637,"to":640,"format":"bf16","rows":"n","dim":"kv_heads*head_dim","fused":false},{"id":"l27.attn.v","from":638,"to":642,"format":"bf16","rows":"n","dim":"kv_heads*head_dim","fused":false},{"id":"l27.attn.v","from":638,"to":643,"format":"bf16","rows":"n","dim":"kv_heads*head_dim","fused":false},{"id":"l27.attn.qn","from":639,"to":641,"format":"bf16","rows":"n","dim":"q_heads*head_dim","fused":false},{"id":"l27.attn.kn","from":640,"to":641,"format":"bf16","rows":"n","dim":"kv_heads*head_dim","fused":false},{"id":"l27.attn.qr","from":641,"to":643,"format":"bf16","rows":"n","dim":"q_heads*head_dim","fused":false},{"id":"l27.attn.kr","from":641,"to":642,"format":"bf16","rows":"n","dim":"kv_heads*head_dim","fused":false},{"id":"l27.attn.kr","from":641,"to":643,"format":"bf16","rows":"n","dim":"kv_heads*head_dim","fused":false},{"id":"l27.attn.a","from":643,"to":644,"format":"bf16","rows":"n","dim":"q_heads*head_dim","fused":false},{"id":"l27.attn.ag","from":644,"to":645,"format":"bf16","rows":"n","dim":"q_heads*head_dim","fused":false},{"id":"l27.attn.o","from":645,"to":646,"format":"bf16","rows":"n","dim":"hidden","fused":false},{"id":"l27.attn.y","from":646,"to":647,"format":"bf16","rows":"n","dim":"hidden","fused":false},{"id":"l27.attn.y","from":646,"to":658,"format":"bf16","rows":"n","dim":"hidden","fused":false},{"id":"l27.moe_ffn.xn","from":647,"to":648,"format":"bf16","rows":"n","dim":"hidden","fused":false},{"id":"l27.moe_ffn.xn","from":647,"to":650,"format":"bf16","rows":"n","dim":"hidden","fused":false},{"id":"l27.moe_ffn.xn","from":647,"to":651,"format":"bf16","rows":"n","dim":"hidden","fused":false},{"id":"l27.moe_ffn.xn","from":647,"to":656,"format":"bf16","rows":"n","dim":"hidden","fused":false},{"id":"l27.moe_ffn.router_logits","from":648,"to":649,"format":"bf16","rows":"n","dim":"experts","fused":false},{"id":"l27.moe_ffn.topk_w","from":649,"to":657,"format":"f32","rows":"n","dim":"top_k","fused":false},{"id":"l27.moe_ffn.topk_id","from":649,"to":650,"format":"i32","rows":"n","dim":"top_k","fused":false},{"id":"l27.moe_ffn.topk_id","from":649,"to":653,"format":"i32","rows":"n","dim":"top_k","fused":false},{"id":"l27.moe_ffn.egu","from":650,"to":652,"format":"bf16","rows":"n*top_k","dim":"moe_inter*2","fused":false},{"id":"l27.moe_ffn.sgu","from":651,"to":654,"format":"bf16","rows":"n","dim":"shared_inter*2","fused":false},{"id":"l27.moe_ffn.eact","from":652,"to":653,"format":"f32","rows":"n*top_k","dim":"moe_inter","fused":true},{"id":"l27.moe_ffn.edown","from":653,"to":657,"format":"bf16","rows":"n*top_k","dim":"hidden","fused":false},{"id":"l27.moe_ffn.sact","from":654,"to":655,"format":"f32","rows":"n","dim":"shared_inter","fused":true},{"id":"l27.moe_ffn.sdown","from":655,"to":657,"format":"bf16","rows":"n","dim":"hidden","fused":false},{"id":"l27.moe_ffn.sg","from":656,"to":657,"format":"f32","rows":"n","dim":"1","fused":true},{"id":"l27.moe_ffn.f","from":657,"to":658,"format":"bf16","rows":"n","dim":"hidden","fused":false},{"id":"l27.moe_ffn.y","from":658,"to":659,"format":"bf16","rows":"n","dim":"hidden","fused":false},{"id":"l27.moe_ffn.y","from":658,"to":669,"format":"bf16","rows":"n","dim":"hidden","fused":false},{"id":"l28.gdn.xn","from":659,"to":660,"format":"bf16","rows":"n","dim":"hidden","fused":false},{"id":"l28.gdn.xn","from":659,"to":661,"format":"bf16","rows":"n","dim":"hidden","fused":false},{"id":"l28.gdn.qkvz","from":660,"to":663,"format":"bf16","rows":"n","dim":"lin_k_heads*lin_k_dim*2+lin_v_heads*lin_v_dim*2","fused":false},{"id":"l28.gdn.qkvz","from":660,"to":667,"format":"bf16","rows":"n","dim":"lin_k_heads*lin_k_dim*2+lin_v_heads*lin_v_dim*2","fused":false},{"id":"l28.gdn.ba","from":661,"to":662,"format":"f32","rows":"n","dim":"lin_v_heads*2","fused":true},{"id":"l28.gdn.decay","from":662,"to":666,"format":"f32","rows":"n","dim":"lin_v_heads","fused":false},{"id":"l28.gdn.beta","from":662,"to":666,"format":"f32","rows":"n","dim":"lin_v_heads","fused":false},{"id":"l28.gdn.conv","from":663,"to":664,"format":"f32","rows":"n","dim":"lin_k_heads*lin_k_dim*2+lin_v_heads*lin_v_dim","fused":true},{"id":"l28.gdn.conv","from":663,"to":665,"format":"f32","rows":"n","dim":"lin_k_heads*lin_k_dim*2+lin_v_heads*lin_v_dim","fused":true},{"id":"l28.gdn.qkv","from":665,"to":666,"format":"f32","rows":"n","dim":"lin_k_heads*lin_k_dim*2+lin_v_heads*lin_v_dim","fused":false},{"id":"l28.gdn.core","from":666,"to":667,"format":"f32","rows":"n","dim":"lin_v_heads*lin_v_dim","fused":false},{"id":"l28.gdn.gated","from":667,"to":668,"format":"bf16","rows":"n","dim":"lin_v_heads*lin_v_dim","fused":false},{"id":"l28.gdn.o","from":668,"to":669,"format":"bf16","rows":"n","dim":"hidden","fused":false},{"id":"l28.gdn.y","from":669,"to":670,"format":"bf16","rows":"n","dim":"hidden","fused":false},{"id":"l28.gdn.y","from":669,"to":681,"format":"bf16","rows":"n","dim":"hidden","fused":false},{"id":"l28.moe_ffn.xn","from":670,"to":671,"format":"bf16","rows":"n","dim":"hidden","fused":false},{"id":"l28.moe_ffn.xn","from":670,"to":673,"format":"bf16","rows":"n","dim":"hidden","fused":false},{"id":"l28.moe_ffn.xn","from":670,"to":674,"format":"bf16","rows":"n","dim":"hidden","fused":false},{"id":"l28.moe_ffn.xn","from":670,"to":679,"format":"bf16","rows":"n","dim":"hidden","fused":false},{"id":"l28.moe_ffn.router_logits","from":671,"to":672,"format":"bf16","rows":"n","dim":"experts","fused":false},{"id":"l28.moe_ffn.topk_w","from":672,"to":680,"format":"f32","rows":"n","dim":"top_k","fused":false},{"id":"l28.moe_ffn.topk_id","from":672,"to":673,"format":"i32","rows":"n","dim":"top_k","fused":false},{"id":"l28.moe_ffn.topk_id","from":672,"to":676,"format":"i32","rows":"n","dim":"top_k","fused":false},{"id":"l28.moe_ffn.egu","from":673,"to":675,"format":"bf16","rows":"n*top_k","dim":"moe_inter*2","fused":false},{"id":"l28.moe_ffn.sgu","from":674,"to":677,"format":"bf16","rows":"n","dim":"shared_inter*2","fused":false},{"id":"l28.moe_ffn.eact","from":675,"to":676,"format":"f32","rows":"n*top_k","dim":"moe_inter","fused":true},{"id":"l28.moe_ffn.edown","from":676,"to":680,"format":"bf16","rows":"n*top_k","dim":"hidden","fused":false},{"id":"l28.moe_ffn.sact","from":677,"to":678,"format":"f32","rows":"n","dim":"shared_inter","fused":true},{"id":"l28.moe_ffn.sdown","from":678,"to":680,"format":"bf16","rows":"n","dim":"hidden","fused":false},{"id":"l28.moe_ffn.sg","from":679,"to":680,"format":"f32","rows":"n","dim":"1","fused":true},{"id":"l28.moe_ffn.f","from":680,"to":681,"format":"bf16","rows":"n","dim":"hidden","fused":false},{"id":"l28.moe_ffn.y","from":681,"to":682,"format":"bf16","rows":"n","dim":"hidden","fused":false},{"id":"l28.moe_ffn.y","from":681,"to":692,"format":"bf16","rows":"n","dim":"hidden","fused":false},{"id":"l29.gdn.xn","from":682,"to":683,"format":"bf16","rows":"n","dim":"hidden","fused":false},{"id":"l29.gdn.xn","from":682,"to":684,"format":"bf16","rows":"n","dim":"hidden","fused":false},{"id":"l29.gdn.qkvz","from":683,"to":686,"format":"bf16","rows":"n","dim":"lin_k_heads*lin_k_dim*2+lin_v_heads*lin_v_dim*2","fused":false},{"id":"l29.gdn.qkvz","from":683,"to":690,"format":"bf16","rows":"n","dim":"lin_k_heads*lin_k_dim*2+lin_v_heads*lin_v_dim*2","fused":false},{"id":"l29.gdn.ba","from":684,"to":685,"format":"f32","rows":"n","dim":"lin_v_heads*2","fused":true},{"id":"l29.gdn.decay","from":685,"to":689,"format":"f32","rows":"n","dim":"lin_v_heads","fused":false},{"id":"l29.gdn.beta","from":685,"to":689,"format":"f32","rows":"n","dim":"lin_v_heads","fused":false},{"id":"l29.gdn.conv","from":686,"to":687,"format":"f32","rows":"n","dim":"lin_k_heads*lin_k_dim*2+lin_v_heads*lin_v_dim","fused":true},{"id":"l29.gdn.conv","from":686,"to":688,"format":"f32","rows":"n","dim":"lin_k_heads*lin_k_dim*2+lin_v_heads*lin_v_dim","fused":true},{"id":"l29.gdn.qkv","from":688,"to":689,"format":"f32","rows":"n","dim":"lin_k_heads*lin_k_dim*2+lin_v_heads*lin_v_dim","fused":false},{"id":"l29.gdn.core","from":689,"to":690,"format":"f32","rows":"n","dim":"lin_v_heads*lin_v_dim","fused":false},{"id":"l29.gdn.gated","from":690,"to":691,"format":"bf16","rows":"n","dim":"lin_v_heads*lin_v_dim","fused":false},{"id":"l29.gdn.o","from":691,"to":692,"format":"bf16","rows":"n","dim":"hidden","fused":false},{"id":"l29.gdn.y","from":692,"to":693,"format":"bf16","rows":"n","dim":"hidden","fused":false},{"id":"l29.gdn.y","from":692,"to":704,"format":"bf16","rows":"n","dim":"hidden","fused":false},{"id":"l29.moe_ffn.xn","from":693,"to":694,"format":"bf16","rows":"n","dim":"hidden","fused":false},{"id":"l29.moe_ffn.xn","from":693,"to":696,"format":"bf16","rows":"n","dim":"hidden","fused":false},{"id":"l29.moe_ffn.xn","from":693,"to":697,"format":"bf16","rows":"n","dim":"hidden","fused":false},{"id":"l29.moe_ffn.xn","from":693,"to":702,"format":"bf16","rows":"n","dim":"hidden","fused":false},{"id":"l29.moe_ffn.router_logits","from":694,"to":695,"format":"bf16","rows":"n","dim":"experts","fused":false},{"id":"l29.moe_ffn.topk_w","from":695,"to":703,"format":"f32","rows":"n","dim":"top_k","fused":false},{"id":"l29.moe_ffn.topk_id","from":695,"to":696,"format":"i32","rows":"n","dim":"top_k","fused":false},{"id":"l29.moe_ffn.topk_id","from":695,"to":699,"format":"i32","rows":"n","dim":"top_k","fused":false},{"id":"l29.moe_ffn.egu","from":696,"to":698,"format":"bf16","rows":"n*top_k","dim":"moe_inter*2","fused":false},{"id":"l29.moe_ffn.sgu","from":697,"to":700,"format":"bf16","rows":"n","dim":"shared_inter*2","fused":false},{"id":"l29.moe_ffn.eact","from":698,"to":699,"format":"f32","rows":"n*top_k","dim":"moe_inter","fused":true},{"id":"l29.moe_ffn.edown","from":699,"to":703,"format":"bf16","rows":"n*top_k","dim":"hidden","fused":false},{"id":"l29.moe_ffn.sact","from":700,"to":701,"format":"f32","rows":"n","dim":"shared_inter","fused":true},{"id":"l29.moe_ffn.sdown","from":701,"to":703,"format":"bf16","rows":"n","dim":"hidden","fused":false},{"id":"l29.moe_ffn.sg","from":702,"to":703,"format":"f32","rows":"n","dim":"1","fused":true},{"id":"l29.moe_ffn.f","from":703,"to":704,"format":"bf16","rows":"n","dim":"hidden","fused":false},{"id":"l29.moe_ffn.y","from":704,"to":705,"format":"bf16","rows":"n","dim":"hidden","fused":false},{"id":"l29.moe_ffn.y","from":704,"to":715,"format":"bf16","rows":"n","dim":"hidden","fused":false},{"id":"l30.gdn.xn","from":705,"to":706,"format":"bf16","rows":"n","dim":"hidden","fused":false},{"id":"l30.gdn.xn","from":705,"to":707,"format":"bf16","rows":"n","dim":"hidden","fused":false},{"id":"l30.gdn.qkvz","from":706,"to":709,"format":"bf16","rows":"n","dim":"lin_k_heads*lin_k_dim*2+lin_v_heads*lin_v_dim*2","fused":false},{"id":"l30.gdn.qkvz","from":706,"to":713,"format":"bf16","rows":"n","dim":"lin_k_heads*lin_k_dim*2+lin_v_heads*lin_v_dim*2","fused":false},{"id":"l30.gdn.ba","from":707,"to":708,"format":"f32","rows":"n","dim":"lin_v_heads*2","fused":true},{"id":"l30.gdn.decay","from":708,"to":712,"format":"f32","rows":"n","dim":"lin_v_heads","fused":false},{"id":"l30.gdn.beta","from":708,"to":712,"format":"f32","rows":"n","dim":"lin_v_heads","fused":false},{"id":"l30.gdn.conv","from":709,"to":710,"format":"f32","rows":"n","dim":"lin_k_heads*lin_k_dim*2+lin_v_heads*lin_v_dim","fused":true},{"id":"l30.gdn.conv","from":709,"to":711,"format":"f32","rows":"n","dim":"lin_k_heads*lin_k_dim*2+lin_v_heads*lin_v_dim","fused":true},{"id":"l30.gdn.qkv","from":711,"to":712,"format":"f32","rows":"n","dim":"lin_k_heads*lin_k_dim*2+lin_v_heads*lin_v_dim","fused":false},{"id":"l30.gdn.core","from":712,"to":713,"format":"f32","rows":"n","dim":"lin_v_heads*lin_v_dim","fused":false},{"id":"l30.gdn.gated","from":713,"to":714,"format":"bf16","rows":"n","dim":"lin_v_heads*lin_v_dim","fused":false},{"id":"l30.gdn.o","from":714,"to":715,"format":"bf16","rows":"n","dim":"hidden","fused":false},{"id":"l30.gdn.y","from":715,"to":716,"format":"bf16","rows":"n","dim":"hidden","fused":false},{"id":"l30.gdn.y","from":715,"to":727,"format":"bf16","rows":"n","dim":"hidden","fused":false},{"id":"l30.moe_ffn.xn","from":716,"to":717,"format":"bf16","rows":"n","dim":"hidden","fused":false},{"id":"l30.moe_ffn.xn","from":716,"to":719,"format":"bf16","rows":"n","dim":"hidden","fused":false},{"id":"l30.moe_ffn.xn","from":716,"to":720,"format":"bf16","rows":"n","dim":"hidden","fused":false},{"id":"l30.moe_ffn.xn","from":716,"to":725,"format":"bf16","rows":"n","dim":"hidden","fused":false},{"id":"l30.moe_ffn.router_logits","from":717,"to":718,"format":"bf16","rows":"n","dim":"experts","fused":false},{"id":"l30.moe_ffn.topk_w","from":718,"to":726,"format":"f32","rows":"n","dim":"top_k","fused":false},{"id":"l30.moe_ffn.topk_id","from":718,"to":719,"format":"i32","rows":"n","dim":"top_k","fused":false},{"id":"l30.moe_ffn.topk_id","from":718,"to":722,"format":"i32","rows":"n","dim":"top_k","fused":false},{"id":"l30.moe_ffn.egu","from":719,"to":721,"format":"bf16","rows":"n*top_k","dim":"moe_inter*2","fused":false},{"id":"l30.moe_ffn.sgu","from":720,"to":723,"format":"bf16","rows":"n","dim":"shared_inter*2","fused":false},{"id":"l30.moe_ffn.eact","from":721,"to":722,"format":"f32","rows":"n*top_k","dim":"moe_inter","fused":true},{"id":"l30.moe_ffn.edown","from":722,"to":726,"format":"bf16","rows":"n*top_k","dim":"hidden","fused":false},{"id":"l30.moe_ffn.sact","from":723,"to":724,"format":"f32","rows":"n","dim":"shared_inter","fused":true},{"id":"l30.moe_ffn.sdown","from":724,"to":726,"format":"bf16","rows":"n","dim":"hidden","fused":false},{"id":"l30.moe_ffn.sg","from":725,"to":726,"format":"f32","rows":"n","dim":"1","fused":true},{"id":"l30.moe_ffn.f","from":726,"to":727,"format":"bf16","rows":"n","dim":"hidden","fused":false},{"id":"l30.moe_ffn.y","from":727,"to":728,"format":"bf16","rows":"n","dim":"hidden","fused":false},{"id":"l30.moe_ffn.y","from":727,"to":740,"format":"bf16","rows":"n","dim":"hidden","fused":false},{"id":"l31.attn.xn","from":728,"to":729,"format":"bf16","rows":"n","dim":"hidden","fused":false},{"id":"l31.attn.xn","from":728,"to":731,"format":"bf16","rows":"n","dim":"hidden","fused":false},{"id":"l31.attn.xn","from":728,"to":732,"format":"bf16","rows":"n","dim":"hidden","fused":false},{"id":"l31.attn.qg","from":729,"to":730,"format":"bf16","rows":"n","dim":"q_heads*head_dim*2","fused":false},{"id":"l31.attn.q","from":730,"to":733,"format":"bf16","rows":"n","dim":"q_heads*head_dim","fused":false},{"id":"l31.attn.gate","from":730,"to":738,"format":"bf16","rows":"n","dim":"q_heads*head_dim","fused":false},{"id":"l31.attn.k","from":731,"to":734,"format":"bf16","rows":"n","dim":"kv_heads*head_dim","fused":false},{"id":"l31.attn.v","from":732,"to":736,"format":"bf16","rows":"n","dim":"kv_heads*head_dim","fused":false},{"id":"l31.attn.v","from":732,"to":737,"format":"bf16","rows":"n","dim":"kv_heads*head_dim","fused":false},{"id":"l31.attn.qn","from":733,"to":735,"format":"bf16","rows":"n","dim":"q_heads*head_dim","fused":false},{"id":"l31.attn.kn","from":734,"to":735,"format":"bf16","rows":"n","dim":"kv_heads*head_dim","fused":false},{"id":"l31.attn.qr","from":735,"to":737,"format":"bf16","rows":"n","dim":"q_heads*head_dim","fused":false},{"id":"l31.attn.kr","from":735,"to":736,"format":"bf16","rows":"n","dim":"kv_heads*head_dim","fused":false},{"id":"l31.attn.kr","from":735,"to":737,"format":"bf16","rows":"n","dim":"kv_heads*head_dim","fused":false},{"id":"l31.attn.a","from":737,"to":738,"format":"bf16","rows":"n","dim":"q_heads*head_dim","fused":false},{"id":"l31.attn.ag","from":738,"to":739,"format":"bf16","rows":"n","dim":"q_heads*head_dim","fused":false},{"id":"l31.attn.o","from":739,"to":740,"format":"bf16","rows":"n","dim":"hidden","fused":false},{"id":"l31.attn.y","from":740,"to":741,"format":"bf16","rows":"n","dim":"hidden","fused":false},{"id":"l31.attn.y","from":740,"to":752,"format":"bf16","rows":"n","dim":"hidden","fused":false},{"id":"l31.moe_ffn.xn","from":741,"to":742,"format":"bf16","rows":"n","dim":"hidden","fused":false},{"id":"l31.moe_ffn.xn","from":741,"to":744,"format":"bf16","rows":"n","dim":"hidden","fused":false},{"id":"l31.moe_ffn.xn","from":741,"to":745,"format":"bf16","rows":"n","dim":"hidden","fused":false},{"id":"l31.moe_ffn.xn","from":741,"to":750,"format":"bf16","rows":"n","dim":"hidden","fused":false},{"id":"l31.moe_ffn.router_logits","from":742,"to":743,"format":"bf16","rows":"n","dim":"experts","fused":false},{"id":"l31.moe_ffn.topk_w","from":743,"to":751,"format":"f32","rows":"n","dim":"top_k","fused":false},{"id":"l31.moe_ffn.topk_id","from":743,"to":744,"format":"i32","rows":"n","dim":"top_k","fused":false},{"id":"l31.moe_ffn.topk_id","from":743,"to":747,"format":"i32","rows":"n","dim":"top_k","fused":false},{"id":"l31.moe_ffn.egu","from":744,"to":746,"format":"bf16","rows":"n*top_k","dim":"moe_inter*2","fused":false},{"id":"l31.moe_ffn.sgu","from":745,"to":748,"format":"bf16","rows":"n","dim":"shared_inter*2","fused":false},{"id":"l31.moe_ffn.eact","from":746,"to":747,"format":"f32","rows":"n*top_k","dim":"moe_inter","fused":true},{"id":"l31.moe_ffn.edown","from":747,"to":751,"format":"bf16","rows":"n*top_k","dim":"hidden","fused":false},{"id":"l31.moe_ffn.sact","from":748,"to":749,"format":"f32","rows":"n","dim":"shared_inter","fused":true},{"id":"l31.moe_ffn.sdown","from":749,"to":751,"format":"bf16","rows":"n","dim":"hidden","fused":false},{"id":"l31.moe_ffn.sg","from":750,"to":751,"format":"f32","rows":"n","dim":"1","fused":true},{"id":"l31.moe_ffn.f","from":751,"to":752,"format":"bf16","rows":"n","dim":"hidden","fused":false},{"id":"l31.moe_ffn.y","from":752,"to":753,"format":"bf16","rows":"n","dim":"hidden","fused":false},{"id":"l31.moe_ffn.y","from":752,"to":763,"format":"bf16","rows":"n","dim":"hidden","fused":false},{"id":"l32.gdn.xn","from":753,"to":754,"format":"bf16","rows":"n","dim":"hidden","fused":false},{"id":"l32.gdn.xn","from":753,"to":755,"format":"bf16","rows":"n","dim":"hidden","fused":false},{"id":"l32.gdn.qkvz","from":754,"to":757,"format":"bf16","rows":"n","dim":"lin_k_heads*lin_k_dim*2+lin_v_heads*lin_v_dim*2","fused":false},{"id":"l32.gdn.qkvz","from":754,"to":761,"format":"bf16","rows":"n","dim":"lin_k_heads*lin_k_dim*2+lin_v_heads*lin_v_dim*2","fused":false},{"id":"l32.gdn.ba","from":755,"to":756,"format":"f32","rows":"n","dim":"lin_v_heads*2","fused":true},{"id":"l32.gdn.decay","from":756,"to":760,"format":"f32","rows":"n","dim":"lin_v_heads","fused":false},{"id":"l32.gdn.beta","from":756,"to":760,"format":"f32","rows":"n","dim":"lin_v_heads","fused":false},{"id":"l32.gdn.conv","from":757,"to":758,"format":"f32","rows":"n","dim":"lin_k_heads*lin_k_dim*2+lin_v_heads*lin_v_dim","fused":true},{"id":"l32.gdn.conv","from":757,"to":759,"format":"f32","rows":"n","dim":"lin_k_heads*lin_k_dim*2+lin_v_heads*lin_v_dim","fused":true},{"id":"l32.gdn.qkv","from":759,"to":760,"format":"f32","rows":"n","dim":"lin_k_heads*lin_k_dim*2+lin_v_heads*lin_v_dim","fused":false},{"id":"l32.gdn.core","from":760,"to":761,"format":"f32","rows":"n","dim":"lin_v_heads*lin_v_dim","fused":false},{"id":"l32.gdn.gated","from":761,"to":762,"format":"bf16","rows":"n","dim":"lin_v_heads*lin_v_dim","fused":false},{"id":"l32.gdn.o","from":762,"to":763,"format":"bf16","rows":"n","dim":"hidden","fused":false},{"id":"l32.gdn.y","from":763,"to":764,"format":"bf16","rows":"n","dim":"hidden","fused":false},{"id":"l32.gdn.y","from":763,"to":775,"format":"bf16","rows":"n","dim":"hidden","fused":false},{"id":"l32.moe_ffn.xn","from":764,"to":765,"format":"bf16","rows":"n","dim":"hidden","fused":false},{"id":"l32.moe_ffn.xn","from":764,"to":767,"format":"bf16","rows":"n","dim":"hidden","fused":false},{"id":"l32.moe_ffn.xn","from":764,"to":768,"format":"bf16","rows":"n","dim":"hidden","fused":false},{"id":"l32.moe_ffn.xn","from":764,"to":773,"format":"bf16","rows":"n","dim":"hidden","fused":false},{"id":"l32.moe_ffn.router_logits","from":765,"to":766,"format":"bf16","rows":"n","dim":"experts","fused":false},{"id":"l32.moe_ffn.topk_w","from":766,"to":774,"format":"f32","rows":"n","dim":"top_k","fused":false},{"id":"l32.moe_ffn.topk_id","from":766,"to":767,"format":"i32","rows":"n","dim":"top_k","fused":false},{"id":"l32.moe_ffn.topk_id","from":766,"to":770,"format":"i32","rows":"n","dim":"top_k","fused":false},{"id":"l32.moe_ffn.egu","from":767,"to":769,"format":"bf16","rows":"n*top_k","dim":"moe_inter*2","fused":false},{"id":"l32.moe_ffn.sgu","from":768,"to":771,"format":"bf16","rows":"n","dim":"shared_inter*2","fused":false},{"id":"l32.moe_ffn.eact","from":769,"to":770,"format":"f32","rows":"n*top_k","dim":"moe_inter","fused":true},{"id":"l32.moe_ffn.edown","from":770,"to":774,"format":"bf16","rows":"n*top_k","dim":"hidden","fused":false},{"id":"l32.moe_ffn.sact","from":771,"to":772,"format":"f32","rows":"n","dim":"shared_inter","fused":true},{"id":"l32.moe_ffn.sdown","from":772,"to":774,"format":"bf16","rows":"n","dim":"hidden","fused":false},{"id":"l32.moe_ffn.sg","from":773,"to":774,"format":"f32","rows":"n","dim":"1","fused":true},{"id":"l32.moe_ffn.f","from":774,"to":775,"format":"bf16","rows":"n","dim":"hidden","fused":false},{"id":"l32.moe_ffn.y","from":775,"to":776,"format":"bf16","rows":"n","dim":"hidden","fused":false},{"id":"l32.moe_ffn.y","from":775,"to":786,"format":"bf16","rows":"n","dim":"hidden","fused":false},{"id":"l33.gdn.xn","from":776,"to":777,"format":"bf16","rows":"n","dim":"hidden","fused":false},{"id":"l33.gdn.xn","from":776,"to":778,"format":"bf16","rows":"n","dim":"hidden","fused":false},{"id":"l33.gdn.qkvz","from":777,"to":780,"format":"bf16","rows":"n","dim":"lin_k_heads*lin_k_dim*2+lin_v_heads*lin_v_dim*2","fused":false},{"id":"l33.gdn.qkvz","from":777,"to":784,"format":"bf16","rows":"n","dim":"lin_k_heads*lin_k_dim*2+lin_v_heads*lin_v_dim*2","fused":false},{"id":"l33.gdn.ba","from":778,"to":779,"format":"f32","rows":"n","dim":"lin_v_heads*2","fused":true},{"id":"l33.gdn.decay","from":779,"to":783,"format":"f32","rows":"n","dim":"lin_v_heads","fused":false},{"id":"l33.gdn.beta","from":779,"to":783,"format":"f32","rows":"n","dim":"lin_v_heads","fused":false},{"id":"l33.gdn.conv","from":780,"to":781,"format":"f32","rows":"n","dim":"lin_k_heads*lin_k_dim*2+lin_v_heads*lin_v_dim","fused":true},{"id":"l33.gdn.conv","from":780,"to":782,"format":"f32","rows":"n","dim":"lin_k_heads*lin_k_dim*2+lin_v_heads*lin_v_dim","fused":true},{"id":"l33.gdn.qkv","from":782,"to":783,"format":"f32","rows":"n","dim":"lin_k_heads*lin_k_dim*2+lin_v_heads*lin_v_dim","fused":false},{"id":"l33.gdn.core","from":783,"to":784,"format":"f32","rows":"n","dim":"lin_v_heads*lin_v_dim","fused":false},{"id":"l33.gdn.gated","from":784,"to":785,"format":"bf16","rows":"n","dim":"lin_v_heads*lin_v_dim","fused":false},{"id":"l33.gdn.o","from":785,"to":786,"format":"bf16","rows":"n","dim":"hidden","fused":false},{"id":"l33.gdn.y","from":786,"to":787,"format":"bf16","rows":"n","dim":"hidden","fused":false},{"id":"l33.gdn.y","from":786,"to":798,"format":"bf16","rows":"n","dim":"hidden","fused":false},{"id":"l33.moe_ffn.xn","from":787,"to":788,"format":"bf16","rows":"n","dim":"hidden","fused":false},{"id":"l33.moe_ffn.xn","from":787,"to":790,"format":"bf16","rows":"n","dim":"hidden","fused":false},{"id":"l33.moe_ffn.xn","from":787,"to":791,"format":"bf16","rows":"n","dim":"hidden","fused":false},{"id":"l33.moe_ffn.xn","from":787,"to":796,"format":"bf16","rows":"n","dim":"hidden","fused":false},{"id":"l33.moe_ffn.router_logits","from":788,"to":789,"format":"bf16","rows":"n","dim":"experts","fused":false},{"id":"l33.moe_ffn.topk_w","from":789,"to":797,"format":"f32","rows":"n","dim":"top_k","fused":false},{"id":"l33.moe_ffn.topk_id","from":789,"to":790,"format":"i32","rows":"n","dim":"top_k","fused":false},{"id":"l33.moe_ffn.topk_id","from":789,"to":793,"format":"i32","rows":"n","dim":"top_k","fused":false},{"id":"l33.moe_ffn.egu","from":790,"to":792,"format":"bf16","rows":"n*top_k","dim":"moe_inter*2","fused":false},{"id":"l33.moe_ffn.sgu","from":791,"to":794,"format":"bf16","rows":"n","dim":"shared_inter*2","fused":false},{"id":"l33.moe_ffn.eact","from":792,"to":793,"format":"f32","rows":"n*top_k","dim":"moe_inter","fused":true},{"id":"l33.moe_ffn.edown","from":793,"to":797,"format":"bf16","rows":"n*top_k","dim":"hidden","fused":false},{"id":"l33.moe_ffn.sact","from":794,"to":795,"format":"f32","rows":"n","dim":"shared_inter","fused":true},{"id":"l33.moe_ffn.sdown","from":795,"to":797,"format":"bf16","rows":"n","dim":"hidden","fused":false},{"id":"l33.moe_ffn.sg","from":796,"to":797,"format":"f32","rows":"n","dim":"1","fused":true},{"id":"l33.moe_ffn.f","from":797,"to":798,"format":"bf16","rows":"n","dim":"hidden","fused":false},{"id":"l33.moe_ffn.y","from":798,"to":799,"format":"bf16","rows":"n","dim":"hidden","fused":false},{"id":"l33.moe_ffn.y","from":798,"to":809,"format":"bf16","rows":"n","dim":"hidden","fused":false},{"id":"l34.gdn.xn","from":799,"to":800,"format":"bf16","rows":"n","dim":"hidden","fused":false},{"id":"l34.gdn.xn","from":799,"to":801,"format":"bf16","rows":"n","dim":"hidden","fused":false},{"id":"l34.gdn.qkvz","from":800,"to":803,"format":"bf16","rows":"n","dim":"lin_k_heads*lin_k_dim*2+lin_v_heads*lin_v_dim*2","fused":false},{"id":"l34.gdn.qkvz","from":800,"to":807,"format":"bf16","rows":"n","dim":"lin_k_heads*lin_k_dim*2+lin_v_heads*lin_v_dim*2","fused":false},{"id":"l34.gdn.ba","from":801,"to":802,"format":"f32","rows":"n","dim":"lin_v_heads*2","fused":true},{"id":"l34.gdn.decay","from":802,"to":806,"format":"f32","rows":"n","dim":"lin_v_heads","fused":false},{"id":"l34.gdn.beta","from":802,"to":806,"format":"f32","rows":"n","dim":"lin_v_heads","fused":false},{"id":"l34.gdn.conv","from":803,"to":804,"format":"f32","rows":"n","dim":"lin_k_heads*lin_k_dim*2+lin_v_heads*lin_v_dim","fused":true},{"id":"l34.gdn.conv","from":803,"to":805,"format":"f32","rows":"n","dim":"lin_k_heads*lin_k_dim*2+lin_v_heads*lin_v_dim","fused":true},{"id":"l34.gdn.qkv","from":805,"to":806,"format":"f32","rows":"n","dim":"lin_k_heads*lin_k_dim*2+lin_v_heads*lin_v_dim","fused":false},{"id":"l34.gdn.core","from":806,"to":807,"format":"f32","rows":"n","dim":"lin_v_heads*lin_v_dim","fused":false},{"id":"l34.gdn.gated","from":807,"to":808,"format":"bf16","rows":"n","dim":"lin_v_heads*lin_v_dim","fused":false},{"id":"l34.gdn.o","from":808,"to":809,"format":"bf16","rows":"n","dim":"hidden","fused":false},{"id":"l34.gdn.y","from":809,"to":810,"format":"bf16","rows":"n","dim":"hidden","fused":false},{"id":"l34.gdn.y","from":809,"to":821,"format":"bf16","rows":"n","dim":"hidden","fused":false},{"id":"l34.moe_ffn.xn","from":810,"to":811,"format":"bf16","rows":"n","dim":"hidden","fused":false},{"id":"l34.moe_ffn.xn","from":810,"to":813,"format":"bf16","rows":"n","dim":"hidden","fused":false},{"id":"l34.moe_ffn.xn","from":810,"to":814,"format":"bf16","rows":"n","dim":"hidden","fused":false},{"id":"l34.moe_ffn.xn","from":810,"to":819,"format":"bf16","rows":"n","dim":"hidden","fused":false},{"id":"l34.moe_ffn.router_logits","from":811,"to":812,"format":"bf16","rows":"n","dim":"experts","fused":false},{"id":"l34.moe_ffn.topk_w","from":812,"to":820,"format":"f32","rows":"n","dim":"top_k","fused":false},{"id":"l34.moe_ffn.topk_id","from":812,"to":813,"format":"i32","rows":"n","dim":"top_k","fused":false},{"id":"l34.moe_ffn.topk_id","from":812,"to":816,"format":"i32","rows":"n","dim":"top_k","fused":false},{"id":"l34.moe_ffn.egu","from":813,"to":815,"format":"bf16","rows":"n*top_k","dim":"moe_inter*2","fused":false},{"id":"l34.moe_ffn.sgu","from":814,"to":817,"format":"bf16","rows":"n","dim":"shared_inter*2","fused":false},{"id":"l34.moe_ffn.eact","from":815,"to":816,"format":"f32","rows":"n*top_k","dim":"moe_inter","fused":true},{"id":"l34.moe_ffn.edown","from":816,"to":820,"format":"bf16","rows":"n*top_k","dim":"hidden","fused":false},{"id":"l34.moe_ffn.sact","from":817,"to":818,"format":"f32","rows":"n","dim":"shared_inter","fused":true},{"id":"l34.moe_ffn.sdown","from":818,"to":820,"format":"bf16","rows":"n","dim":"hidden","fused":false},{"id":"l34.moe_ffn.sg","from":819,"to":820,"format":"f32","rows":"n","dim":"1","fused":true},{"id":"l34.moe_ffn.f","from":820,"to":821,"format":"bf16","rows":"n","dim":"hidden","fused":false},{"id":"l34.moe_ffn.y","from":821,"to":822,"format":"bf16","rows":"n","dim":"hidden","fused":false},{"id":"l34.moe_ffn.y","from":821,"to":834,"format":"bf16","rows":"n","dim":"hidden","fused":false},{"id":"l35.attn.xn","from":822,"to":823,"format":"bf16","rows":"n","dim":"hidden","fused":false},{"id":"l35.attn.xn","from":822,"to":825,"format":"bf16","rows":"n","dim":"hidden","fused":false},{"id":"l35.attn.xn","from":822,"to":826,"format":"bf16","rows":"n","dim":"hidden","fused":false},{"id":"l35.attn.qg","from":823,"to":824,"format":"bf16","rows":"n","dim":"q_heads*head_dim*2","fused":false},{"id":"l35.attn.q","from":824,"to":827,"format":"bf16","rows":"n","dim":"q_heads*head_dim","fused":false},{"id":"l35.attn.gate","from":824,"to":832,"format":"bf16","rows":"n","dim":"q_heads*head_dim","fused":false},{"id":"l35.attn.k","from":825,"to":828,"format":"bf16","rows":"n","dim":"kv_heads*head_dim","fused":false},{"id":"l35.attn.v","from":826,"to":830,"format":"bf16","rows":"n","dim":"kv_heads*head_dim","fused":false},{"id":"l35.attn.v","from":826,"to":831,"format":"bf16","rows":"n","dim":"kv_heads*head_dim","fused":false},{"id":"l35.attn.qn","from":827,"to":829,"format":"bf16","rows":"n","dim":"q_heads*head_dim","fused":false},{"id":"l35.attn.kn","from":828,"to":829,"format":"bf16","rows":"n","dim":"kv_heads*head_dim","fused":false},{"id":"l35.attn.qr","from":829,"to":831,"format":"bf16","rows":"n","dim":"q_heads*head_dim","fused":false},{"id":"l35.attn.kr","from":829,"to":830,"format":"bf16","rows":"n","dim":"kv_heads*head_dim","fused":false},{"id":"l35.attn.kr","from":829,"to":831,"format":"bf16","rows":"n","dim":"kv_heads*head_dim","fused":false},{"id":"l35.attn.a","from":831,"to":832,"format":"bf16","rows":"n","dim":"q_heads*head_dim","fused":false},{"id":"l35.attn.ag","from":832,"to":833,"format":"bf16","rows":"n","dim":"q_heads*head_dim","fused":false},{"id":"l35.attn.o","from":833,"to":834,"format":"bf16","rows":"n","dim":"hidden","fused":false},{"id":"l35.attn.y","from":834,"to":835,"format":"bf16","rows":"n","dim":"hidden","fused":false},{"id":"l35.attn.y","from":834,"to":846,"format":"bf16","rows":"n","dim":"hidden","fused":false},{"id":"l35.moe_ffn.xn","from":835,"to":836,"format":"bf16","rows":"n","dim":"hidden","fused":false},{"id":"l35.moe_ffn.xn","from":835,"to":838,"format":"bf16","rows":"n","dim":"hidden","fused":false},{"id":"l35.moe_ffn.xn","from":835,"to":839,"format":"bf16","rows":"n","dim":"hidden","fused":false},{"id":"l35.moe_ffn.xn","from":835,"to":844,"format":"bf16","rows":"n","dim":"hidden","fused":false},{"id":"l35.moe_ffn.router_logits","from":836,"to":837,"format":"bf16","rows":"n","dim":"experts","fused":false},{"id":"l35.moe_ffn.topk_w","from":837,"to":845,"format":"f32","rows":"n","dim":"top_k","fused":false},{"id":"l35.moe_ffn.topk_id","from":837,"to":838,"format":"i32","rows":"n","dim":"top_k","fused":false},{"id":"l35.moe_ffn.topk_id","from":837,"to":841,"format":"i32","rows":"n","dim":"top_k","fused":false},{"id":"l35.moe_ffn.egu","from":838,"to":840,"format":"bf16","rows":"n*top_k","dim":"moe_inter*2","fused":false},{"id":"l35.moe_ffn.sgu","from":839,"to":842,"format":"bf16","rows":"n","dim":"shared_inter*2","fused":false},{"id":"l35.moe_ffn.eact","from":840,"to":841,"format":"f32","rows":"n*top_k","dim":"moe_inter","fused":true},{"id":"l35.moe_ffn.edown","from":841,"to":845,"format":"bf16","rows":"n*top_k","dim":"hidden","fused":false},{"id":"l35.moe_ffn.sact","from":842,"to":843,"format":"f32","rows":"n","dim":"shared_inter","fused":true},{"id":"l35.moe_ffn.sdown","from":843,"to":845,"format":"bf16","rows":"n","dim":"hidden","fused":false},{"id":"l35.moe_ffn.sg","from":844,"to":845,"format":"f32","rows":"n","dim":"1","fused":true},{"id":"l35.moe_ffn.f","from":845,"to":846,"format":"bf16","rows":"n","dim":"hidden","fused":false},{"id":"l35.moe_ffn.y","from":846,"to":847,"format":"bf16","rows":"n","dim":"hidden","fused":false},{"id":"l35.moe_ffn.y","from":846,"to":857,"format":"bf16","rows":"n","dim":"hidden","fused":false},{"id":"l36.gdn.xn","from":847,"to":848,"format":"bf16","rows":"n","dim":"hidden","fused":false},{"id":"l36.gdn.xn","from":847,"to":849,"format":"bf16","rows":"n","dim":"hidden","fused":false},{"id":"l36.gdn.qkvz","from":848,"to":851,"format":"bf16","rows":"n","dim":"lin_k_heads*lin_k_dim*2+lin_v_heads*lin_v_dim*2","fused":false},{"id":"l36.gdn.qkvz","from":848,"to":855,"format":"bf16","rows":"n","dim":"lin_k_heads*lin_k_dim*2+lin_v_heads*lin_v_dim*2","fused":false},{"id":"l36.gdn.ba","from":849,"to":850,"format":"f32","rows":"n","dim":"lin_v_heads*2","fused":true},{"id":"l36.gdn.decay","from":850,"to":854,"format":"f32","rows":"n","dim":"lin_v_heads","fused":false},{"id":"l36.gdn.beta","from":850,"to":854,"format":"f32","rows":"n","dim":"lin_v_heads","fused":false},{"id":"l36.gdn.conv","from":851,"to":852,"format":"f32","rows":"n","dim":"lin_k_heads*lin_k_dim*2+lin_v_heads*lin_v_dim","fused":true},{"id":"l36.gdn.conv","from":851,"to":853,"format":"f32","rows":"n","dim":"lin_k_heads*lin_k_dim*2+lin_v_heads*lin_v_dim","fused":true},{"id":"l36.gdn.qkv","from":853,"to":854,"format":"f32","rows":"n","dim":"lin_k_heads*lin_k_dim*2+lin_v_heads*lin_v_dim","fused":false},{"id":"l36.gdn.core","from":854,"to":855,"format":"f32","rows":"n","dim":"lin_v_heads*lin_v_dim","fused":false},{"id":"l36.gdn.gated","from":855,"to":856,"format":"bf16","rows":"n","dim":"lin_v_heads*lin_v_dim","fused":false},{"id":"l36.gdn.o","from":856,"to":857,"format":"bf16","rows":"n","dim":"hidden","fused":false},{"id":"l36.gdn.y","from":857,"to":858,"format":"bf16","rows":"n","dim":"hidden","fused":false},{"id":"l36.gdn.y","from":857,"to":869,"format":"bf16","rows":"n","dim":"hidden","fused":false},{"id":"l36.moe_ffn.xn","from":858,"to":859,"format":"bf16","rows":"n","dim":"hidden","fused":false},{"id":"l36.moe_ffn.xn","from":858,"to":861,"format":"bf16","rows":"n","dim":"hidden","fused":false},{"id":"l36.moe_ffn.xn","from":858,"to":862,"format":"bf16","rows":"n","dim":"hidden","fused":false},{"id":"l36.moe_ffn.xn","from":858,"to":867,"format":"bf16","rows":"n","dim":"hidden","fused":false},{"id":"l36.moe_ffn.router_logits","from":859,"to":860,"format":"bf16","rows":"n","dim":"experts","fused":false},{"id":"l36.moe_ffn.topk_w","from":860,"to":868,"format":"f32","rows":"n","dim":"top_k","fused":false},{"id":"l36.moe_ffn.topk_id","from":860,"to":861,"format":"i32","rows":"n","dim":"top_k","fused":false},{"id":"l36.moe_ffn.topk_id","from":860,"to":864,"format":"i32","rows":"n","dim":"top_k","fused":false},{"id":"l36.moe_ffn.egu","from":861,"to":863,"format":"bf16","rows":"n*top_k","dim":"moe_inter*2","fused":false},{"id":"l36.moe_ffn.sgu","from":862,"to":865,"format":"bf16","rows":"n","dim":"shared_inter*2","fused":false},{"id":"l36.moe_ffn.eact","from":863,"to":864,"format":"f32","rows":"n*top_k","dim":"moe_inter","fused":true},{"id":"l36.moe_ffn.edown","from":864,"to":868,"format":"bf16","rows":"n*top_k","dim":"hidden","fused":false},{"id":"l36.moe_ffn.sact","from":865,"to":866,"format":"f32","rows":"n","dim":"shared_inter","fused":true},{"id":"l36.moe_ffn.sdown","from":866,"to":868,"format":"bf16","rows":"n","dim":"hidden","fused":false},{"id":"l36.moe_ffn.sg","from":867,"to":868,"format":"f32","rows":"n","dim":"1","fused":true},{"id":"l36.moe_ffn.f","from":868,"to":869,"format":"bf16","rows":"n","dim":"hidden","fused":false},{"id":"l36.moe_ffn.y","from":869,"to":870,"format":"bf16","rows":"n","dim":"hidden","fused":false},{"id":"l36.moe_ffn.y","from":869,"to":880,"format":"bf16","rows":"n","dim":"hidden","fused":false},{"id":"l37.gdn.xn","from":870,"to":871,"format":"bf16","rows":"n","dim":"hidden","fused":false},{"id":"l37.gdn.xn","from":870,"to":872,"format":"bf16","rows":"n","dim":"hidden","fused":false},{"id":"l37.gdn.qkvz","from":871,"to":874,"format":"bf16","rows":"n","dim":"lin_k_heads*lin_k_dim*2+lin_v_heads*lin_v_dim*2","fused":false},{"id":"l37.gdn.qkvz","from":871,"to":878,"format":"bf16","rows":"n","dim":"lin_k_heads*lin_k_dim*2+lin_v_heads*lin_v_dim*2","fused":false},{"id":"l37.gdn.ba","from":872,"to":873,"format":"f32","rows":"n","dim":"lin_v_heads*2","fused":true},{"id":"l37.gdn.decay","from":873,"to":877,"format":"f32","rows":"n","dim":"lin_v_heads","fused":false},{"id":"l37.gdn.beta","from":873,"to":877,"format":"f32","rows":"n","dim":"lin_v_heads","fused":false},{"id":"l37.gdn.conv","from":874,"to":875,"format":"f32","rows":"n","dim":"lin_k_heads*lin_k_dim*2+lin_v_heads*lin_v_dim","fused":true},{"id":"l37.gdn.conv","from":874,"to":876,"format":"f32","rows":"n","dim":"lin_k_heads*lin_k_dim*2+lin_v_heads*lin_v_dim","fused":true},{"id":"l37.gdn.qkv","from":876,"to":877,"format":"f32","rows":"n","dim":"lin_k_heads*lin_k_dim*2+lin_v_heads*lin_v_dim","fused":false},{"id":"l37.gdn.core","from":877,"to":878,"format":"f32","rows":"n","dim":"lin_v_heads*lin_v_dim","fused":false},{"id":"l37.gdn.gated","from":878,"to":879,"format":"bf16","rows":"n","dim":"lin_v_heads*lin_v_dim","fused":false},{"id":"l37.gdn.o","from":879,"to":880,"format":"bf16","rows":"n","dim":"hidden","fused":false},{"id":"l37.gdn.y","from":880,"to":881,"format":"bf16","rows":"n","dim":"hidden","fused":false},{"id":"l37.gdn.y","from":880,"to":892,"format":"bf16","rows":"n","dim":"hidden","fused":false},{"id":"l37.moe_ffn.xn","from":881,"to":882,"format":"bf16","rows":"n","dim":"hidden","fused":false},{"id":"l37.moe_ffn.xn","from":881,"to":884,"format":"bf16","rows":"n","dim":"hidden","fused":false},{"id":"l37.moe_ffn.xn","from":881,"to":885,"format":"bf16","rows":"n","dim":"hidden","fused":false},{"id":"l37.moe_ffn.xn","from":881,"to":890,"format":"bf16","rows":"n","dim":"hidden","fused":false},{"id":"l37.moe_ffn.router_logits","from":882,"to":883,"format":"bf16","rows":"n","dim":"experts","fused":false},{"id":"l37.moe_ffn.topk_w","from":883,"to":891,"format":"f32","rows":"n","dim":"top_k","fused":false},{"id":"l37.moe_ffn.topk_id","from":883,"to":884,"format":"i32","rows":"n","dim":"top_k","fused":false},{"id":"l37.moe_ffn.topk_id","from":883,"to":887,"format":"i32","rows":"n","dim":"top_k","fused":false},{"id":"l37.moe_ffn.egu","from":884,"to":886,"format":"bf16","rows":"n*top_k","dim":"moe_inter*2","fused":false},{"id":"l37.moe_ffn.sgu","from":885,"to":888,"format":"bf16","rows":"n","dim":"shared_inter*2","fused":false},{"id":"l37.moe_ffn.eact","from":886,"to":887,"format":"f32","rows":"n*top_k","dim":"moe_inter","fused":true},{"id":"l37.moe_ffn.edown","from":887,"to":891,"format":"bf16","rows":"n*top_k","dim":"hidden","fused":false},{"id":"l37.moe_ffn.sact","from":888,"to":889,"format":"f32","rows":"n","dim":"shared_inter","fused":true},{"id":"l37.moe_ffn.sdown","from":889,"to":891,"format":"bf16","rows":"n","dim":"hidden","fused":false},{"id":"l37.moe_ffn.sg","from":890,"to":891,"format":"f32","rows":"n","dim":"1","fused":true},{"id":"l37.moe_ffn.f","from":891,"to":892,"format":"bf16","rows":"n","dim":"hidden","fused":false},{"id":"l37.moe_ffn.y","from":892,"to":893,"format":"bf16","rows":"n","dim":"hidden","fused":false},{"id":"l37.moe_ffn.y","from":892,"to":903,"format":"bf16","rows":"n","dim":"hidden","fused":false},{"id":"l38.gdn.xn","from":893,"to":894,"format":"bf16","rows":"n","dim":"hidden","fused":false},{"id":"l38.gdn.xn","from":893,"to":895,"format":"bf16","rows":"n","dim":"hidden","fused":false},{"id":"l38.gdn.qkvz","from":894,"to":897,"format":"bf16","rows":"n","dim":"lin_k_heads*lin_k_dim*2+lin_v_heads*lin_v_dim*2","fused":false},{"id":"l38.gdn.qkvz","from":894,"to":901,"format":"bf16","rows":"n","dim":"lin_k_heads*lin_k_dim*2+lin_v_heads*lin_v_dim*2","fused":false},{"id":"l38.gdn.ba","from":895,"to":896,"format":"f32","rows":"n","dim":"lin_v_heads*2","fused":true},{"id":"l38.gdn.decay","from":896,"to":900,"format":"f32","rows":"n","dim":"lin_v_heads","fused":false},{"id":"l38.gdn.beta","from":896,"to":900,"format":"f32","rows":"n","dim":"lin_v_heads","fused":false},{"id":"l38.gdn.conv","from":897,"to":898,"format":"f32","rows":"n","dim":"lin_k_heads*lin_k_dim*2+lin_v_heads*lin_v_dim","fused":true},{"id":"l38.gdn.conv","from":897,"to":899,"format":"f32","rows":"n","dim":"lin_k_heads*lin_k_dim*2+lin_v_heads*lin_v_dim","fused":true},{"id":"l38.gdn.qkv","from":899,"to":900,"format":"f32","rows":"n","dim":"lin_k_heads*lin_k_dim*2+lin_v_heads*lin_v_dim","fused":false},{"id":"l38.gdn.core","from":900,"to":901,"format":"f32","rows":"n","dim":"lin_v_heads*lin_v_dim","fused":false},{"id":"l38.gdn.gated","from":901,"to":902,"format":"bf16","rows":"n","dim":"lin_v_heads*lin_v_dim","fused":false},{"id":"l38.gdn.o","from":902,"to":903,"format":"bf16","rows":"n","dim":"hidden","fused":false},{"id":"l38.gdn.y","from":903,"to":904,"format":"bf16","rows":"n","dim":"hidden","fused":false},{"id":"l38.gdn.y","from":903,"to":915,"format":"bf16","rows":"n","dim":"hidden","fused":false},{"id":"l38.moe_ffn.xn","from":904,"to":905,"format":"bf16","rows":"n","dim":"hidden","fused":false},{"id":"l38.moe_ffn.xn","from":904,"to":907,"format":"bf16","rows":"n","dim":"hidden","fused":false},{"id":"l38.moe_ffn.xn","from":904,"to":908,"format":"bf16","rows":"n","dim":"hidden","fused":false},{"id":"l38.moe_ffn.xn","from":904,"to":913,"format":"bf16","rows":"n","dim":"hidden","fused":false},{"id":"l38.moe_ffn.router_logits","from":905,"to":906,"format":"bf16","rows":"n","dim":"experts","fused":false},{"id":"l38.moe_ffn.topk_w","from":906,"to":914,"format":"f32","rows":"n","dim":"top_k","fused":false},{"id":"l38.moe_ffn.topk_id","from":906,"to":907,"format":"i32","rows":"n","dim":"top_k","fused":false},{"id":"l38.moe_ffn.topk_id","from":906,"to":910,"format":"i32","rows":"n","dim":"top_k","fused":false},{"id":"l38.moe_ffn.egu","from":907,"to":909,"format":"bf16","rows":"n*top_k","dim":"moe_inter*2","fused":false},{"id":"l38.moe_ffn.sgu","from":908,"to":911,"format":"bf16","rows":"n","dim":"shared_inter*2","fused":false},{"id":"l38.moe_ffn.eact","from":909,"to":910,"format":"f32","rows":"n*top_k","dim":"moe_inter","fused":true},{"id":"l38.moe_ffn.edown","from":910,"to":914,"format":"bf16","rows":"n*top_k","dim":"hidden","fused":false},{"id":"l38.moe_ffn.sact","from":911,"to":912,"format":"f32","rows":"n","dim":"shared_inter","fused":true},{"id":"l38.moe_ffn.sdown","from":912,"to":914,"format":"bf16","rows":"n","dim":"hidden","fused":false},{"id":"l38.moe_ffn.sg","from":913,"to":914,"format":"f32","rows":"n","dim":"1","fused":true},{"id":"l38.moe_ffn.f","from":914,"to":915,"format":"bf16","rows":"n","dim":"hidden","fused":false},{"id":"l38.moe_ffn.y","from":915,"to":916,"format":"bf16","rows":"n","dim":"hidden","fused":false},{"id":"l38.moe_ffn.y","from":915,"to":928,"format":"bf16","rows":"n","dim":"hidden","fused":false},{"id":"l39.attn.xn","from":916,"to":917,"format":"bf16","rows":"n","dim":"hidden","fused":false},{"id":"l39.attn.xn","from":916,"to":919,"format":"bf16","rows":"n","dim":"hidden","fused":false},{"id":"l39.attn.xn","from":916,"to":920,"format":"bf16","rows":"n","dim":"hidden","fused":false},{"id":"l39.attn.qg","from":917,"to":918,"format":"bf16","rows":"n","dim":"q_heads*head_dim*2","fused":false},{"id":"l39.attn.q","from":918,"to":921,"format":"bf16","rows":"n","dim":"q_heads*head_dim","fused":false},{"id":"l39.attn.gate","from":918,"to":926,"format":"bf16","rows":"n","dim":"q_heads*head_dim","fused":false},{"id":"l39.attn.k","from":919,"to":922,"format":"bf16","rows":"n","dim":"kv_heads*head_dim","fused":false},{"id":"l39.attn.v","from":920,"to":924,"format":"bf16","rows":"n","dim":"kv_heads*head_dim","fused":false},{"id":"l39.attn.v","from":920,"to":925,"format":"bf16","rows":"n","dim":"kv_heads*head_dim","fused":false},{"id":"l39.attn.qn","from":921,"to":923,"format":"bf16","rows":"n","dim":"q_heads*head_dim","fused":false},{"id":"l39.attn.kn","from":922,"to":923,"format":"bf16","rows":"n","dim":"kv_heads*head_dim","fused":false},{"id":"l39.attn.qr","from":923,"to":925,"format":"bf16","rows":"n","dim":"q_heads*head_dim","fused":false},{"id":"l39.attn.kr","from":923,"to":924,"format":"bf16","rows":"n","dim":"kv_heads*head_dim","fused":false},{"id":"l39.attn.kr","from":923,"to":925,"format":"bf16","rows":"n","dim":"kv_heads*head_dim","fused":false},{"id":"l39.attn.a","from":925,"to":926,"format":"bf16","rows":"n","dim":"q_heads*head_dim","fused":false},{"id":"l39.attn.ag","from":926,"to":927,"format":"bf16","rows":"n","dim":"q_heads*head_dim","fused":false},{"id":"l39.attn.o","from":927,"to":928,"format":"bf16","rows":"n","dim":"hidden","fused":false},{"id":"l39.attn.y","from":928,"to":929,"format":"bf16","rows":"n","dim":"hidden","fused":false},{"id":"l39.attn.y","from":928,"to":940,"format":"bf16","rows":"n","dim":"hidden","fused":false},{"id":"l39.moe_ffn.xn","from":929,"to":930,"format":"bf16","rows":"n","dim":"hidden","fused":false},{"id":"l39.moe_ffn.xn","from":929,"to":932,"format":"bf16","rows":"n","dim":"hidden","fused":false},{"id":"l39.moe_ffn.xn","from":929,"to":933,"format":"bf16","rows":"n","dim":"hidden","fused":false},{"id":"l39.moe_ffn.xn","from":929,"to":938,"format":"bf16","rows":"n","dim":"hidden","fused":false},{"id":"l39.moe_ffn.router_logits","from":930,"to":931,"format":"bf16","rows":"n","dim":"experts","fused":false},{"id":"l39.moe_ffn.topk_w","from":931,"to":939,"format":"f32","rows":"n","dim":"top_k","fused":false},{"id":"l39.moe_ffn.topk_id","from":931,"to":932,"format":"i32","rows":"n","dim":"top_k","fused":false},{"id":"l39.moe_ffn.topk_id","from":931,"to":935,"format":"i32","rows":"n","dim":"top_k","fused":false},{"id":"l39.moe_ffn.egu","from":932,"to":934,"format":"bf16","rows":"n*top_k","dim":"moe_inter*2","fused":false},{"id":"l39.moe_ffn.sgu","from":933,"to":936,"format":"bf16","rows":"n","dim":"shared_inter*2","fused":false},{"id":"l39.moe_ffn.eact","from":934,"to":935,"format":"f32","rows":"n*top_k","dim":"moe_inter","fused":true},{"id":"l39.moe_ffn.edown","from":935,"to":939,"format":"bf16","rows":"n*top_k","dim":"hidden","fused":false},{"id":"l39.moe_ffn.sact","from":936,"to":937,"format":"f32","rows":"n","dim":"shared_inter","fused":true},{"id":"l39.moe_ffn.sdown","from":937,"to":939,"format":"bf16","rows":"n","dim":"hidden","fused":false},{"id":"l39.moe_ffn.sg","from":938,"to":939,"format":"f32","rows":"n","dim":"1","fused":true},{"id":"l39.moe_ffn.f","from":939,"to":940,"format":"bf16","rows":"n","dim":"hidden","fused":false},{"id":"l39.moe_ffn.y","from":940,"to":941,"format":"bf16","rows":"n","dim":"hidden","fused":false},{"id":"head.xn","from":941,"to":942,"format":"bf16","rows":"n","dim":"hidden","fused":false},{"id":"head.logits","from":942,"to":943,"format":"bf16","rows":"n","dim":"vocab","fused":false}],"groups":[{"id":0,"rule":"embed_row_copy","kernels":[],"where":"host","numerics":"reference","compute":null,"nodes":[0]},{"id":1,"rule":"input_norm_residual","kernels":["norm::rms_norm_residual"],"where":"device","numerics":"reference","compute":"memory","nodes":[1]},{"id":2,"rule":"w8a16_m32","kernels":["w8a16_gemm_pipelined_m32::w8a16_gemm_pipelined_m32"],"where":"device","numerics":"reference","compute":"tensor_core:mma.sync.m16n8k16.bf16","nodes":[2]},{"id":3,"rule":"gdn_ba_gates_gemv_per_row","kernels":["ssm_preprocess::dense_gemv_ba_gates"],"where":"device","numerics":"reference","compute":"cuda_core","nodes":[3,4]},{"id":4,"rule":"gdn_conv_l2_f32_per_row","kernels":["causal_conv1d::causal_conv1d_update_l2norm_f32"],"where":"device","numerics":"reference","compute":"cuda_core","nodes":[5,6,7]},{"id":5,"rule":"gdn_recurrence_f32_per_row","kernels":["gated_delta_rule::gated_delta_rule_decode_f32"],"where":"device","numerics":"reference","compute":"cuda_core","nodes":[8]},{"id":6,"rule":"gdn_out_norm_f32_per_row","kernels":["norm::gated_rms_norm_f32_input"],"where":"device","numerics":"reference","compute":"memory","nodes":[9]},{"id":7,"rule":"w8a16_m32","kernels":["w8a16_gemm_pipelined_m32::w8a16_gemm_pipelined_m32"],"where":"device","numerics":"reference","compute":"tensor_core:mma.sync.m16n8k16.bf16","nodes":[10]},{"id":8,"rule":"residual_add_post_norm","kernels":["norm::residual_add_rms_norm"],"where":"device","numerics":"reference","compute":"memory","nodes":[11,12]},{"id":9,"rule":"moe_router_gemv","kernels":["gemv::dense_gemv_bf16"],"where":"device","numerics":"reference","compute":"cuda_core","nodes":[13]},{"id":10,"rule":"moe_topk_softmax","kernels":["moe_topk::moe_topk_softmax"],"where":"device","numerics":"reference","compute":"memory","nodes":[14]},{"id":11,"rule":"moe_gate_up_shared_fp8","kernels":["moe_shared_expert_fused_fp8::moe_expert_gate_up_shared_fp8"],"where":"device","numerics":"reference","compute":"cuda_core","nodes":[15,16]},{"id":12,"rule":"moe_silu_down_shared_fp8","kernels":["moe_shared_expert_fused_fp8::moe_expert_silu_down_shared_fp8"],"where":"device","numerics":"reference","compute":"cuda_core","nodes":[17,18,19,20]},{"id":13,"rule":"moe_weighted_sum_blend","kernels":["moe_expert_gemv::moe_weighted_sum_blend"],"where":"device","numerics":"reference","compute":"cuda_core","nodes":[21,22]},{"id":14,"rule":"cross_layer_add_norm","kernels":["residual_add_rms_norm_exact::residual_add_rms_norm_exact"],"where":"device","numerics":"bit_identical(residual_add_rms_norm_exact_microtest)","compute":"memory","nodes":[23,24]},{"id":15,"rule":"w8a16_m32","kernels":["w8a16_gemm_pipelined_m32::w8a16_gemm_pipelined_m32"],"where":"device","numerics":"reference","compute":"tensor_core:mma.sync.m16n8k16.bf16","nodes":[25]},{"id":16,"rule":"gdn_ba_gates_gemv_per_row","kernels":["ssm_preprocess::dense_gemv_ba_gates"],"where":"device","numerics":"reference","compute":"cuda_core","nodes":[26,27]},{"id":17,"rule":"gdn_conv_l2_f32_per_row","kernels":["causal_conv1d::causal_conv1d_update_l2norm_f32"],"where":"device","numerics":"reference","compute":"cuda_core","nodes":[28,29,30]},{"id":18,"rule":"gdn_recurrence_f32_per_row","kernels":["gated_delta_rule::gated_delta_rule_decode_f32"],"where":"device","numerics":"reference","compute":"cuda_core","nodes":[31]},{"id":19,"rule":"gdn_out_norm_f32_per_row","kernels":["norm::gated_rms_norm_f32_input"],"where":"device","numerics":"reference","compute":"memory","nodes":[32]},{"id":20,"rule":"w8a16_m32","kernels":["w8a16_gemm_pipelined_m32::w8a16_gemm_pipelined_m32"],"where":"device","numerics":"reference","compute":"tensor_core:mma.sync.m16n8k16.bf16","nodes":[33]},{"id":21,"rule":"residual_add_post_norm","kernels":["norm::residual_add_rms_norm"],"where":"device","numerics":"reference","compute":"memory","nodes":[34,35]},{"id":22,"rule":"moe_router_gemv","kernels":["gemv::dense_gemv_bf16"],"where":"device","numerics":"reference","compute":"cuda_core","nodes":[36]},{"id":23,"rule":"moe_topk_softmax","kernels":["moe_topk::moe_topk_softmax"],"where":"device","numerics":"reference","compute":"memory","nodes":[37]},{"id":24,"rule":"moe_gate_up_shared_fp8","kernels":["moe_shared_expert_fused_fp8::moe_expert_gate_up_shared_fp8"],"where":"device","numerics":"reference","compute":"cuda_core","nodes":[38,39]},{"id":25,"rule":"moe_silu_down_shared_fp8","kernels":["moe_shared_expert_fused_fp8::moe_expert_silu_down_shared_fp8"],"where":"device","numerics":"reference","compute":"cuda_core","nodes":[40,41,42,43]},{"id":26,"rule":"moe_weighted_sum_blend","kernels":["moe_expert_gemv::moe_weighted_sum_blend"],"where":"device","numerics":"reference","compute":"cuda_core","nodes":[44,45]},{"id":27,"rule":"cross_layer_add_norm","kernels":["residual_add_rms_norm_exact::residual_add_rms_norm_exact"],"where":"device","numerics":"bit_identical(residual_add_rms_norm_exact_microtest)","compute":"memory","nodes":[46,47]},{"id":28,"rule":"w8a16_m32","kernels":["w8a16_gemm_pipelined_m32::w8a16_gemm_pipelined_m32"],"where":"device","numerics":"reference","compute":"tensor_core:mma.sync.m16n8k16.bf16","nodes":[48]},{"id":29,"rule":"gdn_ba_gates_gemv_per_row","kernels":["ssm_preprocess::dense_gemv_ba_gates"],"where":"device","numerics":"reference","compute":"cuda_core","nodes":[49,50]},{"id":30,"rule":"gdn_conv_l2_f32_per_row","kernels":["causal_conv1d::causal_conv1d_update_l2norm_f32"],"where":"device","numerics":"reference","compute":"cuda_core","nodes":[51,52,53]},{"id":31,"rule":"gdn_recurrence_f32_per_row","kernels":["gated_delta_rule::gated_delta_rule_decode_f32"],"where":"device","numerics":"reference","compute":"cuda_core","nodes":[54]},{"id":32,"rule":"gdn_out_norm_f32_per_row","kernels":["norm::gated_rms_norm_f32_input"],"where":"device","numerics":"reference","compute":"memory","nodes":[55]},{"id":33,"rule":"w8a16_m32","kernels":["w8a16_gemm_pipelined_m32::w8a16_gemm_pipelined_m32"],"where":"device","numerics":"reference","compute":"tensor_core:mma.sync.m16n8k16.bf16","nodes":[56]},{"id":34,"rule":"residual_add_post_norm","kernels":["norm::residual_add_rms_norm"],"where":"device","numerics":"reference","compute":"memory","nodes":[57,58]},{"id":35,"rule":"moe_router_gemv","kernels":["gemv::dense_gemv_bf16"],"where":"device","numerics":"reference","compute":"cuda_core","nodes":[59]},{"id":36,"rule":"moe_topk_softmax","kernels":["moe_topk::moe_topk_softmax"],"where":"device","numerics":"reference","compute":"memory","nodes":[60]},{"id":37,"rule":"moe_gate_up_shared_fp8","kernels":["moe_shared_expert_fused_fp8::moe_expert_gate_up_shared_fp8"],"where":"device","numerics":"reference","compute":"cuda_core","nodes":[61,62]},{"id":38,"rule":"moe_silu_down_shared_fp8","kernels":["moe_shared_expert_fused_fp8::moe_expert_silu_down_shared_fp8"],"where":"device","numerics":"reference","compute":"cuda_core","nodes":[63,64,65,66]},{"id":39,"rule":"moe_weighted_sum_blend","kernels":["moe_expert_gemv::moe_weighted_sum_blend"],"where":"device","numerics":"reference","compute":"cuda_core","nodes":[67,68]},{"id":40,"rule":"cross_layer_add_norm","kernels":["residual_add_rms_norm_exact::residual_add_rms_norm_exact"],"where":"device","numerics":"bit_identical(residual_add_rms_norm_exact_microtest)","compute":"memory","nodes":[69,70]},{"id":41,"rule":"w8a16_m32","kernels":["w8a16_gemm_pipelined_m32::w8a16_gemm_pipelined_m32"],"where":"device","numerics":"reference","compute":"tensor_core:mma.sync.m16n8k16.bf16","nodes":[71]},{"id":42,"rule":"deinterleave_qg","kernels":["ssm_preprocess::deinterleave_qg"],"where":"device","numerics":"reference","compute":"memory","nodes":[72]},{"id":43,"rule":"w8a16_m32","kernels":["w8a16_gemm_pipelined_m32::w8a16_gemm_pipelined_m32"],"where":"device","numerics":"reference","compute":"tensor_core:mma.sync.m16n8k16.bf16","nodes":[73]},{"id":44,"rule":"w8a16_m32","kernels":["w8a16_gemm_pipelined_m32::w8a16_gemm_pipelined_m32"],"where":"device","numerics":"reference","compute":"tensor_core:mma.sync.m16n8k16.bf16","nodes":[74]},{"id":45,"rule":"qk_norm_rows","kernels":["norm::rms_norm"],"where":"device","numerics":"reference","compute":"memory","nodes":[75]},{"id":46,"rule":"qk_norm_rows","kernels":["norm::rms_norm"],"where":"device","numerics":"reference","compute":"memory","nodes":[76]},{"id":47,"rule":"rope_mrope_interleaved","kernels":["rope_mrope_interleaved::rope_forward_mrope_interleaved"],"where":"device","numerics":"reference","compute":"memory","nodes":[77]},{"id":48,"rule":"kv_write_bf16","kernels":["reshape_and_cache::reshape_and_cache_flash"],"where":"device","numerics":"reference","compute":"memory","nodes":[78]},{"id":49,"rule":"paged_attention_bf16","kernels":["paged_decode::paged_decode_attn"],"where":"device","numerics":"reference","compute":"cuda_core","nodes":[79]},{"id":50,"rule":"sigmoid_gate_mul","kernels":["residual_add::sigmoid_gate_mul"],"where":"device","numerics":"reference","compute":"memory","nodes":[80]},{"id":51,"rule":"w8a16_m32","kernels":["w8a16_gemm_pipelined_m32::w8a16_gemm_pipelined_m32"],"where":"device","numerics":"reference","compute":"tensor_core:mma.sync.m16n8k16.bf16","nodes":[81]},{"id":52,"rule":"residual_add_post_norm","kernels":["norm::residual_add_rms_norm"],"where":"device","numerics":"reference","compute":"memory","nodes":[82,83]},{"id":53,"rule":"moe_router_gemv","kernels":["gemv::dense_gemv_bf16"],"where":"device","numerics":"reference","compute":"cuda_core","nodes":[84]},{"id":54,"rule":"moe_topk_softmax","kernels":["moe_topk::moe_topk_softmax"],"where":"device","numerics":"reference","compute":"memory","nodes":[85]},{"id":55,"rule":"moe_gate_up_shared_fp8","kernels":["moe_shared_expert_fused_fp8::moe_expert_gate_up_shared_fp8"],"where":"device","numerics":"reference","compute":"cuda_core","nodes":[86,87]},{"id":56,"rule":"moe_silu_down_shared_fp8","kernels":["moe_shared_expert_fused_fp8::moe_expert_silu_down_shared_fp8"],"where":"device","numerics":"reference","compute":"cuda_core","nodes":[88,89,90,91]},{"id":57,"rule":"moe_weighted_sum_blend","kernels":["moe_expert_gemv::moe_weighted_sum_blend"],"where":"device","numerics":"reference","compute":"cuda_core","nodes":[92,93]},{"id":58,"rule":"cross_layer_add_norm","kernels":["residual_add_rms_norm_exact::residual_add_rms_norm_exact"],"where":"device","numerics":"bit_identical(residual_add_rms_norm_exact_microtest)","compute":"memory","nodes":[94,95]},{"id":59,"rule":"w8a16_m32","kernels":["w8a16_gemm_pipelined_m32::w8a16_gemm_pipelined_m32"],"where":"device","numerics":"reference","compute":"tensor_core:mma.sync.m16n8k16.bf16","nodes":[96]},{"id":60,"rule":"gdn_ba_gates_gemv_per_row","kernels":["ssm_preprocess::dense_gemv_ba_gates"],"where":"device","numerics":"reference","compute":"cuda_core","nodes":[97,98]},{"id":61,"rule":"gdn_conv_l2_f32_per_row","kernels":["causal_conv1d::causal_conv1d_update_l2norm_f32"],"where":"device","numerics":"reference","compute":"cuda_core","nodes":[99,100,101]},{"id":62,"rule":"gdn_recurrence_f32_per_row","kernels":["gated_delta_rule::gated_delta_rule_decode_f32"],"where":"device","numerics":"reference","compute":"cuda_core","nodes":[102]},{"id":63,"rule":"gdn_out_norm_f32_per_row","kernels":["norm::gated_rms_norm_f32_input"],"where":"device","numerics":"reference","compute":"memory","nodes":[103]},{"id":64,"rule":"w8a16_m32","kernels":["w8a16_gemm_pipelined_m32::w8a16_gemm_pipelined_m32"],"where":"device","numerics":"reference","compute":"tensor_core:mma.sync.m16n8k16.bf16","nodes":[104]},{"id":65,"rule":"residual_add_post_norm","kernels":["norm::residual_add_rms_norm"],"where":"device","numerics":"reference","compute":"memory","nodes":[105,106]},{"id":66,"rule":"moe_router_gemv","kernels":["gemv::dense_gemv_bf16"],"where":"device","numerics":"reference","compute":"cuda_core","nodes":[107]},{"id":67,"rule":"moe_topk_softmax","kernels":["moe_topk::moe_topk_softmax"],"where":"device","numerics":"reference","compute":"memory","nodes":[108]},{"id":68,"rule":"moe_gate_up_shared_fp8","kernels":["moe_shared_expert_fused_fp8::moe_expert_gate_up_shared_fp8"],"where":"device","numerics":"reference","compute":"cuda_core","nodes":[109,110]},{"id":69,"rule":"moe_silu_down_shared_fp8","kernels":["moe_shared_expert_fused_fp8::moe_expert_silu_down_shared_fp8"],"where":"device","numerics":"reference","compute":"cuda_core","nodes":[111,112,113,114]},{"id":70,"rule":"moe_weighted_sum_blend","kernels":["moe_expert_gemv::moe_weighted_sum_blend"],"where":"device","numerics":"reference","compute":"cuda_core","nodes":[115,116]},{"id":71,"rule":"cross_layer_add_norm","kernels":["residual_add_rms_norm_exact::residual_add_rms_norm_exact"],"where":"device","numerics":"bit_identical(residual_add_rms_norm_exact_microtest)","compute":"memory","nodes":[117,118]},{"id":72,"rule":"w8a16_m32","kernels":["w8a16_gemm_pipelined_m32::w8a16_gemm_pipelined_m32"],"where":"device","numerics":"reference","compute":"tensor_core:mma.sync.m16n8k16.bf16","nodes":[119]},{"id":73,"rule":"gdn_ba_gates_gemv_per_row","kernels":["ssm_preprocess::dense_gemv_ba_gates"],"where":"device","numerics":"reference","compute":"cuda_core","nodes":[120,121]},{"id":74,"rule":"gdn_conv_l2_f32_per_row","kernels":["causal_conv1d::causal_conv1d_update_l2norm_f32"],"where":"device","numerics":"reference","compute":"cuda_core","nodes":[122,123,124]},{"id":75,"rule":"gdn_recurrence_f32_per_row","kernels":["gated_delta_rule::gated_delta_rule_decode_f32"],"where":"device","numerics":"reference","compute":"cuda_core","nodes":[125]},{"id":76,"rule":"gdn_out_norm_f32_per_row","kernels":["norm::gated_rms_norm_f32_input"],"where":"device","numerics":"reference","compute":"memory","nodes":[126]},{"id":77,"rule":"w8a16_m32","kernels":["w8a16_gemm_pipelined_m32::w8a16_gemm_pipelined_m32"],"where":"device","numerics":"reference","compute":"tensor_core:mma.sync.m16n8k16.bf16","nodes":[127]},{"id":78,"rule":"residual_add_post_norm","kernels":["norm::residual_add_rms_norm"],"where":"device","numerics":"reference","compute":"memory","nodes":[128,129]},{"id":79,"rule":"moe_router_gemv","kernels":["gemv::dense_gemv_bf16"],"where":"device","numerics":"reference","compute":"cuda_core","nodes":[130]},{"id":80,"rule":"moe_topk_softmax","kernels":["moe_topk::moe_topk_softmax"],"where":"device","numerics":"reference","compute":"memory","nodes":[131]},{"id":81,"rule":"moe_gate_up_shared_fp8","kernels":["moe_shared_expert_fused_fp8::moe_expert_gate_up_shared_fp8"],"where":"device","numerics":"reference","compute":"cuda_core","nodes":[132,133]},{"id":82,"rule":"moe_silu_down_shared_fp8","kernels":["moe_shared_expert_fused_fp8::moe_expert_silu_down_shared_fp8"],"where":"device","numerics":"reference","compute":"cuda_core","nodes":[134,135,136,137]},{"id":83,"rule":"moe_weighted_sum_blend","kernels":["moe_expert_gemv::moe_weighted_sum_blend"],"where":"device","numerics":"reference","compute":"cuda_core","nodes":[138,139]},{"id":84,"rule":"cross_layer_add_norm","kernels":["residual_add_rms_norm_exact::residual_add_rms_norm_exact"],"where":"device","numerics":"bit_identical(residual_add_rms_norm_exact_microtest)","compute":"memory","nodes":[140,141]},{"id":85,"rule":"w8a16_m32","kernels":["w8a16_gemm_pipelined_m32::w8a16_gemm_pipelined_m32"],"where":"device","numerics":"reference","compute":"tensor_core:mma.sync.m16n8k16.bf16","nodes":[142]},{"id":86,"rule":"gdn_ba_gates_gemv_per_row","kernels":["ssm_preprocess::dense_gemv_ba_gates"],"where":"device","numerics":"reference","compute":"cuda_core","nodes":[143,144]},{"id":87,"rule":"gdn_conv_l2_f32_per_row","kernels":["causal_conv1d::causal_conv1d_update_l2norm_f32"],"where":"device","numerics":"reference","compute":"cuda_core","nodes":[145,146,147]},{"id":88,"rule":"gdn_recurrence_f32_per_row","kernels":["gated_delta_rule::gated_delta_rule_decode_f32"],"where":"device","numerics":"reference","compute":"cuda_core","nodes":[148]},{"id":89,"rule":"gdn_out_norm_f32_per_row","kernels":["norm::gated_rms_norm_f32_input"],"where":"device","numerics":"reference","compute":"memory","nodes":[149]},{"id":90,"rule":"w8a16_m32","kernels":["w8a16_gemm_pipelined_m32::w8a16_gemm_pipelined_m32"],"where":"device","numerics":"reference","compute":"tensor_core:mma.sync.m16n8k16.bf16","nodes":[150]},{"id":91,"rule":"residual_add_post_norm","kernels":["norm::residual_add_rms_norm"],"where":"device","numerics":"reference","compute":"memory","nodes":[151,152]},{"id":92,"rule":"moe_router_gemv","kernels":["gemv::dense_gemv_bf16"],"where":"device","numerics":"reference","compute":"cuda_core","nodes":[153]},{"id":93,"rule":"moe_topk_softmax","kernels":["moe_topk::moe_topk_softmax"],"where":"device","numerics":"reference","compute":"memory","nodes":[154]},{"id":94,"rule":"moe_gate_up_shared_fp8","kernels":["moe_shared_expert_fused_fp8::moe_expert_gate_up_shared_fp8"],"where":"device","numerics":"reference","compute":"cuda_core","nodes":[155,156]},{"id":95,"rule":"moe_silu_down_shared_fp8","kernels":["moe_shared_expert_fused_fp8::moe_expert_silu_down_shared_fp8"],"where":"device","numerics":"reference","compute":"cuda_core","nodes":[157,158,159,160]},{"id":96,"rule":"moe_weighted_sum_blend","kernels":["moe_expert_gemv::moe_weighted_sum_blend"],"where":"device","numerics":"reference","compute":"cuda_core","nodes":[161,162]},{"id":97,"rule":"cross_layer_add_norm","kernels":["residual_add_rms_norm_exact::residual_add_rms_norm_exact"],"where":"device","numerics":"bit_identical(residual_add_rms_norm_exact_microtest)","compute":"memory","nodes":[163,164]},{"id":98,"rule":"w8a16_m32","kernels":["w8a16_gemm_pipelined_m32::w8a16_gemm_pipelined_m32"],"where":"device","numerics":"reference","compute":"tensor_core:mma.sync.m16n8k16.bf16","nodes":[165]},{"id":99,"rule":"deinterleave_qg","kernels":["ssm_preprocess::deinterleave_qg"],"where":"device","numerics":"reference","compute":"memory","nodes":[166]},{"id":100,"rule":"w8a16_m32","kernels":["w8a16_gemm_pipelined_m32::w8a16_gemm_pipelined_m32"],"where":"device","numerics":"reference","compute":"tensor_core:mma.sync.m16n8k16.bf16","nodes":[167]},{"id":101,"rule":"w8a16_m32","kernels":["w8a16_gemm_pipelined_m32::w8a16_gemm_pipelined_m32"],"where":"device","numerics":"reference","compute":"tensor_core:mma.sync.m16n8k16.bf16","nodes":[168]},{"id":102,"rule":"qk_norm_rows","kernels":["norm::rms_norm"],"where":"device","numerics":"reference","compute":"memory","nodes":[169]},{"id":103,"rule":"qk_norm_rows","kernels":["norm::rms_norm"],"where":"device","numerics":"reference","compute":"memory","nodes":[170]},{"id":104,"rule":"rope_mrope_interleaved","kernels":["rope_mrope_interleaved::rope_forward_mrope_interleaved"],"where":"device","numerics":"reference","compute":"memory","nodes":[171]},{"id":105,"rule":"kv_write_bf16","kernels":["reshape_and_cache::reshape_and_cache_flash"],"where":"device","numerics":"reference","compute":"memory","nodes":[172]},{"id":106,"rule":"paged_attention_bf16","kernels":["paged_decode::paged_decode_attn"],"where":"device","numerics":"reference","compute":"cuda_core","nodes":[173]},{"id":107,"rule":"sigmoid_gate_mul","kernels":["residual_add::sigmoid_gate_mul"],"where":"device","numerics":"reference","compute":"memory","nodes":[174]},{"id":108,"rule":"w8a16_m32","kernels":["w8a16_gemm_pipelined_m32::w8a16_gemm_pipelined_m32"],"where":"device","numerics":"reference","compute":"tensor_core:mma.sync.m16n8k16.bf16","nodes":[175]},{"id":109,"rule":"residual_add_post_norm","kernels":["norm::residual_add_rms_norm"],"where":"device","numerics":"reference","compute":"memory","nodes":[176,177]},{"id":110,"rule":"moe_router_gemv","kernels":["gemv::dense_gemv_bf16"],"where":"device","numerics":"reference","compute":"cuda_core","nodes":[178]},{"id":111,"rule":"moe_topk_softmax","kernels":["moe_topk::moe_topk_softmax"],"where":"device","numerics":"reference","compute":"memory","nodes":[179]},{"id":112,"rule":"moe_gate_up_shared_fp8","kernels":["moe_shared_expert_fused_fp8::moe_expert_gate_up_shared_fp8"],"where":"device","numerics":"reference","compute":"cuda_core","nodes":[180,181]},{"id":113,"rule":"moe_silu_down_shared_fp8","kernels":["moe_shared_expert_fused_fp8::moe_expert_silu_down_shared_fp8"],"where":"device","numerics":"reference","compute":"cuda_core","nodes":[182,183,184,185]},{"id":114,"rule":"moe_weighted_sum_blend","kernels":["moe_expert_gemv::moe_weighted_sum_blend"],"where":"device","numerics":"reference","compute":"cuda_core","nodes":[186,187]},{"id":115,"rule":"cross_layer_add_norm","kernels":["residual_add_rms_norm_exact::residual_add_rms_norm_exact"],"where":"device","numerics":"bit_identical(residual_add_rms_norm_exact_microtest)","compute":"memory","nodes":[188,189]},{"id":116,"rule":"w8a16_m32","kernels":["w8a16_gemm_pipelined_m32::w8a16_gemm_pipelined_m32"],"where":"device","numerics":"reference","compute":"tensor_core:mma.sync.m16n8k16.bf16","nodes":[190]},{"id":117,"rule":"gdn_ba_gates_gemv_per_row","kernels":["ssm_preprocess::dense_gemv_ba_gates"],"where":"device","numerics":"reference","compute":"cuda_core","nodes":[191,192]},{"id":118,"rule":"gdn_conv_l2_f32_per_row","kernels":["causal_conv1d::causal_conv1d_update_l2norm_f32"],"where":"device","numerics":"reference","compute":"cuda_core","nodes":[193,194,195]},{"id":119,"rule":"gdn_recurrence_f32_per_row","kernels":["gated_delta_rule::gated_delta_rule_decode_f32"],"where":"device","numerics":"reference","compute":"cuda_core","nodes":[196]},{"id":120,"rule":"gdn_out_norm_f32_per_row","kernels":["norm::gated_rms_norm_f32_input"],"where":"device","numerics":"reference","compute":"memory","nodes":[197]},{"id":121,"rule":"w8a16_m32","kernels":["w8a16_gemm_pipelined_m32::w8a16_gemm_pipelined_m32"],"where":"device","numerics":"reference","compute":"tensor_core:mma.sync.m16n8k16.bf16","nodes":[198]},{"id":122,"rule":"residual_add_post_norm","kernels":["norm::residual_add_rms_norm"],"where":"device","numerics":"reference","compute":"memory","nodes":[199,200]},{"id":123,"rule":"moe_router_gemv","kernels":["gemv::dense_gemv_bf16"],"where":"device","numerics":"reference","compute":"cuda_core","nodes":[201]},{"id":124,"rule":"moe_topk_softmax","kernels":["moe_topk::moe_topk_softmax"],"where":"device","numerics":"reference","compute":"memory","nodes":[202]},{"id":125,"rule":"moe_gate_up_shared_fp8","kernels":["moe_shared_expert_fused_fp8::moe_expert_gate_up_shared_fp8"],"where":"device","numerics":"reference","compute":"cuda_core","nodes":[203,204]},{"id":126,"rule":"moe_silu_down_shared_fp8","kernels":["moe_shared_expert_fused_fp8::moe_expert_silu_down_shared_fp8"],"where":"device","numerics":"reference","compute":"cuda_core","nodes":[205,206,207,208]},{"id":127,"rule":"moe_weighted_sum_blend","kernels":["moe_expert_gemv::moe_weighted_sum_blend"],"where":"device","numerics":"reference","compute":"cuda_core","nodes":[209,210]},{"id":128,"rule":"cross_layer_add_norm","kernels":["residual_add_rms_norm_exact::residual_add_rms_norm_exact"],"where":"device","numerics":"bit_identical(residual_add_rms_norm_exact_microtest)","compute":"memory","nodes":[211,212]},{"id":129,"rule":"w8a16_m32","kernels":["w8a16_gemm_pipelined_m32::w8a16_gemm_pipelined_m32"],"where":"device","numerics":"reference","compute":"tensor_core:mma.sync.m16n8k16.bf16","nodes":[213]},{"id":130,"rule":"gdn_ba_gates_gemv_per_row","kernels":["ssm_preprocess::dense_gemv_ba_gates"],"where":"device","numerics":"reference","compute":"cuda_core","nodes":[214,215]},{"id":131,"rule":"gdn_conv_l2_f32_per_row","kernels":["causal_conv1d::causal_conv1d_update_l2norm_f32"],"where":"device","numerics":"reference","compute":"cuda_core","nodes":[216,217,218]},{"id":132,"rule":"gdn_recurrence_f32_per_row","kernels":["gated_delta_rule::gated_delta_rule_decode_f32"],"where":"device","numerics":"reference","compute":"cuda_core","nodes":[219]},{"id":133,"rule":"gdn_out_norm_f32_per_row","kernels":["norm::gated_rms_norm_f32_input"],"where":"device","numerics":"reference","compute":"memory","nodes":[220]},{"id":134,"rule":"w8a16_m32","kernels":["w8a16_gemm_pipelined_m32::w8a16_gemm_pipelined_m32"],"where":"device","numerics":"reference","compute":"tensor_core:mma.sync.m16n8k16.bf16","nodes":[221]},{"id":135,"rule":"residual_add_post_norm","kernels":["norm::residual_add_rms_norm"],"where":"device","numerics":"reference","compute":"memory","nodes":[222,223]},{"id":136,"rule":"moe_router_gemv","kernels":["gemv::dense_gemv_bf16"],"where":"device","numerics":"reference","compute":"cuda_core","nodes":[224]},{"id":137,"rule":"moe_topk_softmax","kernels":["moe_topk::moe_topk_softmax"],"where":"device","numerics":"reference","compute":"memory","nodes":[225]},{"id":138,"rule":"moe_gate_up_shared_fp8","kernels":["moe_shared_expert_fused_fp8::moe_expert_gate_up_shared_fp8"],"where":"device","numerics":"reference","compute":"cuda_core","nodes":[226,227]},{"id":139,"rule":"moe_silu_down_shared_fp8","kernels":["moe_shared_expert_fused_fp8::moe_expert_silu_down_shared_fp8"],"where":"device","numerics":"reference","compute":"cuda_core","nodes":[228,229,230,231]},{"id":140,"rule":"moe_weighted_sum_blend","kernels":["moe_expert_gemv::moe_weighted_sum_blend"],"where":"device","numerics":"reference","compute":"cuda_core","nodes":[232,233]},{"id":141,"rule":"cross_layer_add_norm","kernels":["residual_add_rms_norm_exact::residual_add_rms_norm_exact"],"where":"device","numerics":"bit_identical(residual_add_rms_norm_exact_microtest)","compute":"memory","nodes":[234,235]},{"id":142,"rule":"w8a16_m32","kernels":["w8a16_gemm_pipelined_m32::w8a16_gemm_pipelined_m32"],"where":"device","numerics":"reference","compute":"tensor_core:mma.sync.m16n8k16.bf16","nodes":[236]},{"id":143,"rule":"gdn_ba_gates_gemv_per_row","kernels":["ssm_preprocess::dense_gemv_ba_gates"],"where":"device","numerics":"reference","compute":"cuda_core","nodes":[237,238]},{"id":144,"rule":"gdn_conv_l2_f32_per_row","kernels":["causal_conv1d::causal_conv1d_update_l2norm_f32"],"where":"device","numerics":"reference","compute":"cuda_core","nodes":[239,240,241]},{"id":145,"rule":"gdn_recurrence_f32_per_row","kernels":["gated_delta_rule::gated_delta_rule_decode_f32"],"where":"device","numerics":"reference","compute":"cuda_core","nodes":[242]},{"id":146,"rule":"gdn_out_norm_f32_per_row","kernels":["norm::gated_rms_norm_f32_input"],"where":"device","numerics":"reference","compute":"memory","nodes":[243]},{"id":147,"rule":"w8a16_m32","kernels":["w8a16_gemm_pipelined_m32::w8a16_gemm_pipelined_m32"],"where":"device","numerics":"reference","compute":"tensor_core:mma.sync.m16n8k16.bf16","nodes":[244]},{"id":148,"rule":"residual_add_post_norm","kernels":["norm::residual_add_rms_norm"],"where":"device","numerics":"reference","compute":"memory","nodes":[245,246]},{"id":149,"rule":"moe_router_gemv","kernels":["gemv::dense_gemv_bf16"],"where":"device","numerics":"reference","compute":"cuda_core","nodes":[247]},{"id":150,"rule":"moe_topk_softmax","kernels":["moe_topk::moe_topk_softmax"],"where":"device","numerics":"reference","compute":"memory","nodes":[248]},{"id":151,"rule":"moe_gate_up_shared_fp8","kernels":["moe_shared_expert_fused_fp8::moe_expert_gate_up_shared_fp8"],"where":"device","numerics":"reference","compute":"cuda_core","nodes":[249,250]},{"id":152,"rule":"moe_silu_down_shared_fp8","kernels":["moe_shared_expert_fused_fp8::moe_expert_silu_down_shared_fp8"],"where":"device","numerics":"reference","compute":"cuda_core","nodes":[251,252,253,254]},{"id":153,"rule":"moe_weighted_sum_blend","kernels":["moe_expert_gemv::moe_weighted_sum_blend"],"where":"device","numerics":"reference","compute":"cuda_core","nodes":[255,256]},{"id":154,"rule":"cross_layer_add_norm","kernels":["residual_add_rms_norm_exact::residual_add_rms_norm_exact"],"where":"device","numerics":"bit_identical(residual_add_rms_norm_exact_microtest)","compute":"memory","nodes":[257,258]},{"id":155,"rule":"w8a16_m32","kernels":["w8a16_gemm_pipelined_m32::w8a16_gemm_pipelined_m32"],"where":"device","numerics":"reference","compute":"tensor_core:mma.sync.m16n8k16.bf16","nodes":[259]},{"id":156,"rule":"deinterleave_qg","kernels":["ssm_preprocess::deinterleave_qg"],"where":"device","numerics":"reference","compute":"memory","nodes":[260]},{"id":157,"rule":"w8a16_m32","kernels":["w8a16_gemm_pipelined_m32::w8a16_gemm_pipelined_m32"],"where":"device","numerics":"reference","compute":"tensor_core:mma.sync.m16n8k16.bf16","nodes":[261]},{"id":158,"rule":"w8a16_m32","kernels":["w8a16_gemm_pipelined_m32::w8a16_gemm_pipelined_m32"],"where":"device","numerics":"reference","compute":"tensor_core:mma.sync.m16n8k16.bf16","nodes":[262]},{"id":159,"rule":"qk_norm_rows","kernels":["norm::rms_norm"],"where":"device","numerics":"reference","compute":"memory","nodes":[263]},{"id":160,"rule":"qk_norm_rows","kernels":["norm::rms_norm"],"where":"device","numerics":"reference","compute":"memory","nodes":[264]},{"id":161,"rule":"rope_mrope_interleaved","kernels":["rope_mrope_interleaved::rope_forward_mrope_interleaved"],"where":"device","numerics":"reference","compute":"memory","nodes":[265]},{"id":162,"rule":"kv_write_bf16","kernels":["reshape_and_cache::reshape_and_cache_flash"],"where":"device","numerics":"reference","compute":"memory","nodes":[266]},{"id":163,"rule":"paged_attention_bf16","kernels":["paged_decode::paged_decode_attn"],"where":"device","numerics":"reference","compute":"cuda_core","nodes":[267]},{"id":164,"rule":"sigmoid_gate_mul","kernels":["residual_add::sigmoid_gate_mul"],"where":"device","numerics":"reference","compute":"memory","nodes":[268]},{"id":165,"rule":"w8a16_m32","kernels":["w8a16_gemm_pipelined_m32::w8a16_gemm_pipelined_m32"],"where":"device","numerics":"reference","compute":"tensor_core:mma.sync.m16n8k16.bf16","nodes":[269]},{"id":166,"rule":"residual_add_post_norm","kernels":["norm::residual_add_rms_norm"],"where":"device","numerics":"reference","compute":"memory","nodes":[270,271]},{"id":167,"rule":"moe_router_gemv","kernels":["gemv::dense_gemv_bf16"],"where":"device","numerics":"reference","compute":"cuda_core","nodes":[272]},{"id":168,"rule":"moe_topk_softmax","kernels":["moe_topk::moe_topk_softmax"],"where":"device","numerics":"reference","compute":"memory","nodes":[273]},{"id":169,"rule":"moe_gate_up_shared_fp8","kernels":["moe_shared_expert_fused_fp8::moe_expert_gate_up_shared_fp8"],"where":"device","numerics":"reference","compute":"cuda_core","nodes":[274,275]},{"id":170,"rule":"moe_silu_down_shared_fp8","kernels":["moe_shared_expert_fused_fp8::moe_expert_silu_down_shared_fp8"],"where":"device","numerics":"reference","compute":"cuda_core","nodes":[276,277,278,279]},{"id":171,"rule":"moe_weighted_sum_blend","kernels":["moe_expert_gemv::moe_weighted_sum_blend"],"where":"device","numerics":"reference","compute":"cuda_core","nodes":[280,281]},{"id":172,"rule":"cross_layer_add_norm","kernels":["residual_add_rms_norm_exact::residual_add_rms_norm_exact"],"where":"device","numerics":"bit_identical(residual_add_rms_norm_exact_microtest)","compute":"memory","nodes":[282,283]},{"id":173,"rule":"w8a16_m32","kernels":["w8a16_gemm_pipelined_m32::w8a16_gemm_pipelined_m32"],"where":"device","numerics":"reference","compute":"tensor_core:mma.sync.m16n8k16.bf16","nodes":[284]},{"id":174,"rule":"gdn_ba_gates_gemv_per_row","kernels":["ssm_preprocess::dense_gemv_ba_gates"],"where":"device","numerics":"reference","compute":"cuda_core","nodes":[285,286]},{"id":175,"rule":"gdn_conv_l2_f32_per_row","kernels":["causal_conv1d::causal_conv1d_update_l2norm_f32"],"where":"device","numerics":"reference","compute":"cuda_core","nodes":[287,288,289]},{"id":176,"rule":"gdn_recurrence_f32_per_row","kernels":["gated_delta_rule::gated_delta_rule_decode_f32"],"where":"device","numerics":"reference","compute":"cuda_core","nodes":[290]},{"id":177,"rule":"gdn_out_norm_f32_per_row","kernels":["norm::gated_rms_norm_f32_input"],"where":"device","numerics":"reference","compute":"memory","nodes":[291]},{"id":178,"rule":"w8a16_m32","kernels":["w8a16_gemm_pipelined_m32::w8a16_gemm_pipelined_m32"],"where":"device","numerics":"reference","compute":"tensor_core:mma.sync.m16n8k16.bf16","nodes":[292]},{"id":179,"rule":"residual_add_post_norm","kernels":["norm::residual_add_rms_norm"],"where":"device","numerics":"reference","compute":"memory","nodes":[293,294]},{"id":180,"rule":"moe_router_gemv","kernels":["gemv::dense_gemv_bf16"],"where":"device","numerics":"reference","compute":"cuda_core","nodes":[295]},{"id":181,"rule":"moe_topk_softmax","kernels":["moe_topk::moe_topk_softmax"],"where":"device","numerics":"reference","compute":"memory","nodes":[296]},{"id":182,"rule":"moe_gate_up_shared_fp8","kernels":["moe_shared_expert_fused_fp8::moe_expert_gate_up_shared_fp8"],"where":"device","numerics":"reference","compute":"cuda_core","nodes":[297,298]},{"id":183,"rule":"moe_silu_down_shared_fp8","kernels":["moe_shared_expert_fused_fp8::moe_expert_silu_down_shared_fp8"],"where":"device","numerics":"reference","compute":"cuda_core","nodes":[299,300,301,302]},{"id":184,"rule":"moe_weighted_sum_blend","kernels":["moe_expert_gemv::moe_weighted_sum_blend"],"where":"device","numerics":"reference","compute":"cuda_core","nodes":[303,304]},{"id":185,"rule":"cross_layer_add_norm","kernels":["residual_add_rms_norm_exact::residual_add_rms_norm_exact"],"where":"device","numerics":"bit_identical(residual_add_rms_norm_exact_microtest)","compute":"memory","nodes":[305,306]},{"id":186,"rule":"w8a16_m32","kernels":["w8a16_gemm_pipelined_m32::w8a16_gemm_pipelined_m32"],"where":"device","numerics":"reference","compute":"tensor_core:mma.sync.m16n8k16.bf16","nodes":[307]},{"id":187,"rule":"gdn_ba_gates_gemv_per_row","kernels":["ssm_preprocess::dense_gemv_ba_gates"],"where":"device","numerics":"reference","compute":"cuda_core","nodes":[308,309]},{"id":188,"rule":"gdn_conv_l2_f32_per_row","kernels":["causal_conv1d::causal_conv1d_update_l2norm_f32"],"where":"device","numerics":"reference","compute":"cuda_core","nodes":[310,311,312]},{"id":189,"rule":"gdn_recurrence_f32_per_row","kernels":["gated_delta_rule::gated_delta_rule_decode_f32"],"where":"device","numerics":"reference","compute":"cuda_core","nodes":[313]},{"id":190,"rule":"gdn_out_norm_f32_per_row","kernels":["norm::gated_rms_norm_f32_input"],"where":"device","numerics":"reference","compute":"memory","nodes":[314]},{"id":191,"rule":"w8a16_m32","kernels":["w8a16_gemm_pipelined_m32::w8a16_gemm_pipelined_m32"],"where":"device","numerics":"reference","compute":"tensor_core:mma.sync.m16n8k16.bf16","nodes":[315]},{"id":192,"rule":"residual_add_post_norm","kernels":["norm::residual_add_rms_norm"],"where":"device","numerics":"reference","compute":"memory","nodes":[316,317]},{"id":193,"rule":"moe_router_gemv","kernels":["gemv::dense_gemv_bf16"],"where":"device","numerics":"reference","compute":"cuda_core","nodes":[318]},{"id":194,"rule":"moe_topk_softmax","kernels":["moe_topk::moe_topk_softmax"],"where":"device","numerics":"reference","compute":"memory","nodes":[319]},{"id":195,"rule":"moe_gate_up_shared_fp8","kernels":["moe_shared_expert_fused_fp8::moe_expert_gate_up_shared_fp8"],"where":"device","numerics":"reference","compute":"cuda_core","nodes":[320,321]},{"id":196,"rule":"moe_silu_down_shared_fp8","kernels":["moe_shared_expert_fused_fp8::moe_expert_silu_down_shared_fp8"],"where":"device","numerics":"reference","compute":"cuda_core","nodes":[322,323,324,325]},{"id":197,"rule":"moe_weighted_sum_blend","kernels":["moe_expert_gemv::moe_weighted_sum_blend"],"where":"device","numerics":"reference","compute":"cuda_core","nodes":[326,327]},{"id":198,"rule":"cross_layer_add_norm","kernels":["residual_add_rms_norm_exact::residual_add_rms_norm_exact"],"where":"device","numerics":"bit_identical(residual_add_rms_norm_exact_microtest)","compute":"memory","nodes":[328,329]},{"id":199,"rule":"w8a16_m32","kernels":["w8a16_gemm_pipelined_m32::w8a16_gemm_pipelined_m32"],"where":"device","numerics":"reference","compute":"tensor_core:mma.sync.m16n8k16.bf16","nodes":[330]},{"id":200,"rule":"gdn_ba_gates_gemv_per_row","kernels":["ssm_preprocess::dense_gemv_ba_gates"],"where":"device","numerics":"reference","compute":"cuda_core","nodes":[331,332]},{"id":201,"rule":"gdn_conv_l2_f32_per_row","kernels":["causal_conv1d::causal_conv1d_update_l2norm_f32"],"where":"device","numerics":"reference","compute":"cuda_core","nodes":[333,334,335]},{"id":202,"rule":"gdn_recurrence_f32_per_row","kernels":["gated_delta_rule::gated_delta_rule_decode_f32"],"where":"device","numerics":"reference","compute":"cuda_core","nodes":[336]},{"id":203,"rule":"gdn_out_norm_f32_per_row","kernels":["norm::gated_rms_norm_f32_input"],"where":"device","numerics":"reference","compute":"memory","nodes":[337]},{"id":204,"rule":"w8a16_m32","kernels":["w8a16_gemm_pipelined_m32::w8a16_gemm_pipelined_m32"],"where":"device","numerics":"reference","compute":"tensor_core:mma.sync.m16n8k16.bf16","nodes":[338]},{"id":205,"rule":"residual_add_post_norm","kernels":["norm::residual_add_rms_norm"],"where":"device","numerics":"reference","compute":"memory","nodes":[339,340]},{"id":206,"rule":"moe_router_gemv","kernels":["gemv::dense_gemv_bf16"],"where":"device","numerics":"reference","compute":"cuda_core","nodes":[341]},{"id":207,"rule":"moe_topk_softmax","kernels":["moe_topk::moe_topk_softmax"],"where":"device","numerics":"reference","compute":"memory","nodes":[342]},{"id":208,"rule":"moe_gate_up_shared_fp8","kernels":["moe_shared_expert_fused_fp8::moe_expert_gate_up_shared_fp8"],"where":"device","numerics":"reference","compute":"cuda_core","nodes":[343,344]},{"id":209,"rule":"moe_silu_down_shared_fp8","kernels":["moe_shared_expert_fused_fp8::moe_expert_silu_down_shared_fp8"],"where":"device","numerics":"reference","compute":"cuda_core","nodes":[345,346,347,348]},{"id":210,"rule":"moe_weighted_sum_blend","kernels":["moe_expert_gemv::moe_weighted_sum_blend"],"where":"device","numerics":"reference","compute":"cuda_core","nodes":[349,350]},{"id":211,"rule":"cross_layer_add_norm","kernels":["residual_add_rms_norm_exact::residual_add_rms_norm_exact"],"where":"device","numerics":"bit_identical(residual_add_rms_norm_exact_microtest)","compute":"memory","nodes":[351,352]},{"id":212,"rule":"w8a16_m32","kernels":["w8a16_gemm_pipelined_m32::w8a16_gemm_pipelined_m32"],"where":"device","numerics":"reference","compute":"tensor_core:mma.sync.m16n8k16.bf16","nodes":[353]},{"id":213,"rule":"deinterleave_qg","kernels":["ssm_preprocess::deinterleave_qg"],"where":"device","numerics":"reference","compute":"memory","nodes":[354]},{"id":214,"rule":"w8a16_m32","kernels":["w8a16_gemm_pipelined_m32::w8a16_gemm_pipelined_m32"],"where":"device","numerics":"reference","compute":"tensor_core:mma.sync.m16n8k16.bf16","nodes":[355]},{"id":215,"rule":"w8a16_m32","kernels":["w8a16_gemm_pipelined_m32::w8a16_gemm_pipelined_m32"],"where":"device","numerics":"reference","compute":"tensor_core:mma.sync.m16n8k16.bf16","nodes":[356]},{"id":216,"rule":"qk_norm_rows","kernels":["norm::rms_norm"],"where":"device","numerics":"reference","compute":"memory","nodes":[357]},{"id":217,"rule":"qk_norm_rows","kernels":["norm::rms_norm"],"where":"device","numerics":"reference","compute":"memory","nodes":[358]},{"id":218,"rule":"rope_mrope_interleaved","kernels":["rope_mrope_interleaved::rope_forward_mrope_interleaved"],"where":"device","numerics":"reference","compute":"memory","nodes":[359]},{"id":219,"rule":"kv_write_bf16","kernels":["reshape_and_cache::reshape_and_cache_flash"],"where":"device","numerics":"reference","compute":"memory","nodes":[360]},{"id":220,"rule":"paged_attention_bf16","kernels":["paged_decode::paged_decode_attn"],"where":"device","numerics":"reference","compute":"cuda_core","nodes":[361]},{"id":221,"rule":"sigmoid_gate_mul","kernels":["residual_add::sigmoid_gate_mul"],"where":"device","numerics":"reference","compute":"memory","nodes":[362]},{"id":222,"rule":"w8a16_m32","kernels":["w8a16_gemm_pipelined_m32::w8a16_gemm_pipelined_m32"],"where":"device","numerics":"reference","compute":"tensor_core:mma.sync.m16n8k16.bf16","nodes":[363]},{"id":223,"rule":"residual_add_post_norm","kernels":["norm::residual_add_rms_norm"],"where":"device","numerics":"reference","compute":"memory","nodes":[364,365]},{"id":224,"rule":"moe_router_gemv","kernels":["gemv::dense_gemv_bf16"],"where":"device","numerics":"reference","compute":"cuda_core","nodes":[366]},{"id":225,"rule":"moe_topk_softmax","kernels":["moe_topk::moe_topk_softmax"],"where":"device","numerics":"reference","compute":"memory","nodes":[367]},{"id":226,"rule":"moe_gate_up_shared_fp8","kernels":["moe_shared_expert_fused_fp8::moe_expert_gate_up_shared_fp8"],"where":"device","numerics":"reference","compute":"cuda_core","nodes":[368,369]},{"id":227,"rule":"moe_silu_down_shared_fp8","kernels":["moe_shared_expert_fused_fp8::moe_expert_silu_down_shared_fp8"],"where":"device","numerics":"reference","compute":"cuda_core","nodes":[370,371,372,373]},{"id":228,"rule":"moe_weighted_sum_blend","kernels":["moe_expert_gemv::moe_weighted_sum_blend"],"where":"device","numerics":"reference","compute":"cuda_core","nodes":[374,375]},{"id":229,"rule":"cross_layer_add_norm","kernels":["residual_add_rms_norm_exact::residual_add_rms_norm_exact"],"where":"device","numerics":"bit_identical(residual_add_rms_norm_exact_microtest)","compute":"memory","nodes":[376,377]},{"id":230,"rule":"w8a16_m32","kernels":["w8a16_gemm_pipelined_m32::w8a16_gemm_pipelined_m32"],"where":"device","numerics":"reference","compute":"tensor_core:mma.sync.m16n8k16.bf16","nodes":[378]},{"id":231,"rule":"gdn_ba_gates_gemv_per_row","kernels":["ssm_preprocess::dense_gemv_ba_gates"],"where":"device","numerics":"reference","compute":"cuda_core","nodes":[379,380]},{"id":232,"rule":"gdn_conv_l2_f32_per_row","kernels":["causal_conv1d::causal_conv1d_update_l2norm_f32"],"where":"device","numerics":"reference","compute":"cuda_core","nodes":[381,382,383]},{"id":233,"rule":"gdn_recurrence_f32_per_row","kernels":["gated_delta_rule::gated_delta_rule_decode_f32"],"where":"device","numerics":"reference","compute":"cuda_core","nodes":[384]},{"id":234,"rule":"gdn_out_norm_f32_per_row","kernels":["norm::gated_rms_norm_f32_input"],"where":"device","numerics":"reference","compute":"memory","nodes":[385]},{"id":235,"rule":"w8a16_m32","kernels":["w8a16_gemm_pipelined_m32::w8a16_gemm_pipelined_m32"],"where":"device","numerics":"reference","compute":"tensor_core:mma.sync.m16n8k16.bf16","nodes":[386]},{"id":236,"rule":"residual_add_post_norm","kernels":["norm::residual_add_rms_norm"],"where":"device","numerics":"reference","compute":"memory","nodes":[387,388]},{"id":237,"rule":"moe_router_gemv","kernels":["gemv::dense_gemv_bf16"],"where":"device","numerics":"reference","compute":"cuda_core","nodes":[389]},{"id":238,"rule":"moe_topk_softmax","kernels":["moe_topk::moe_topk_softmax"],"where":"device","numerics":"reference","compute":"memory","nodes":[390]},{"id":239,"rule":"moe_gate_up_shared_fp8","kernels":["moe_shared_expert_fused_fp8::moe_expert_gate_up_shared_fp8"],"where":"device","numerics":"reference","compute":"cuda_core","nodes":[391,392]},{"id":240,"rule":"moe_silu_down_shared_fp8","kernels":["moe_shared_expert_fused_fp8::moe_expert_silu_down_shared_fp8"],"where":"device","numerics":"reference","compute":"cuda_core","nodes":[393,394,395,396]},{"id":241,"rule":"moe_weighted_sum_blend","kernels":["moe_expert_gemv::moe_weighted_sum_blend"],"where":"device","numerics":"reference","compute":"cuda_core","nodes":[397,398]},{"id":242,"rule":"cross_layer_add_norm","kernels":["residual_add_rms_norm_exact::residual_add_rms_norm_exact"],"where":"device","numerics":"bit_identical(residual_add_rms_norm_exact_microtest)","compute":"memory","nodes":[399,400]},{"id":243,"rule":"w8a16_m32","kernels":["w8a16_gemm_pipelined_m32::w8a16_gemm_pipelined_m32"],"where":"device","numerics":"reference","compute":"tensor_core:mma.sync.m16n8k16.bf16","nodes":[401]},{"id":244,"rule":"gdn_ba_gates_gemv_per_row","kernels":["ssm_preprocess::dense_gemv_ba_gates"],"where":"device","numerics":"reference","compute":"cuda_core","nodes":[402,403]},{"id":245,"rule":"gdn_conv_l2_f32_per_row","kernels":["causal_conv1d::causal_conv1d_update_l2norm_f32"],"where":"device","numerics":"reference","compute":"cuda_core","nodes":[404,405,406]},{"id":246,"rule":"gdn_recurrence_f32_per_row","kernels":["gated_delta_rule::gated_delta_rule_decode_f32"],"where":"device","numerics":"reference","compute":"cuda_core","nodes":[407]},{"id":247,"rule":"gdn_out_norm_f32_per_row","kernels":["norm::gated_rms_norm_f32_input"],"where":"device","numerics":"reference","compute":"memory","nodes":[408]},{"id":248,"rule":"w8a16_m32","kernels":["w8a16_gemm_pipelined_m32::w8a16_gemm_pipelined_m32"],"where":"device","numerics":"reference","compute":"tensor_core:mma.sync.m16n8k16.bf16","nodes":[409]},{"id":249,"rule":"residual_add_post_norm","kernels":["norm::residual_add_rms_norm"],"where":"device","numerics":"reference","compute":"memory","nodes":[410,411]},{"id":250,"rule":"moe_router_gemv","kernels":["gemv::dense_gemv_bf16"],"where":"device","numerics":"reference","compute":"cuda_core","nodes":[412]},{"id":251,"rule":"moe_topk_softmax","kernels":["moe_topk::moe_topk_softmax"],"where":"device","numerics":"reference","compute":"memory","nodes":[413]},{"id":252,"rule":"moe_gate_up_shared_fp8","kernels":["moe_shared_expert_fused_fp8::moe_expert_gate_up_shared_fp8"],"where":"device","numerics":"reference","compute":"cuda_core","nodes":[414,415]},{"id":253,"rule":"moe_silu_down_shared_fp8","kernels":["moe_shared_expert_fused_fp8::moe_expert_silu_down_shared_fp8"],"where":"device","numerics":"reference","compute":"cuda_core","nodes":[416,417,418,419]},{"id":254,"rule":"moe_weighted_sum_blend","kernels":["moe_expert_gemv::moe_weighted_sum_blend"],"where":"device","numerics":"reference","compute":"cuda_core","nodes":[420,421]},{"id":255,"rule":"cross_layer_add_norm","kernels":["residual_add_rms_norm_exact::residual_add_rms_norm_exact"],"where":"device","numerics":"bit_identical(residual_add_rms_norm_exact_microtest)","compute":"memory","nodes":[422,423]},{"id":256,"rule":"w8a16_m32","kernels":["w8a16_gemm_pipelined_m32::w8a16_gemm_pipelined_m32"],"where":"device","numerics":"reference","compute":"tensor_core:mma.sync.m16n8k16.bf16","nodes":[424]},{"id":257,"rule":"gdn_ba_gates_gemv_per_row","kernels":["ssm_preprocess::dense_gemv_ba_gates"],"where":"device","numerics":"reference","compute":"cuda_core","nodes":[425,426]},{"id":258,"rule":"gdn_conv_l2_f32_per_row","kernels":["causal_conv1d::causal_conv1d_update_l2norm_f32"],"where":"device","numerics":"reference","compute":"cuda_core","nodes":[427,428,429]},{"id":259,"rule":"gdn_recurrence_f32_per_row","kernels":["gated_delta_rule::gated_delta_rule_decode_f32"],"where":"device","numerics":"reference","compute":"cuda_core","nodes":[430]},{"id":260,"rule":"gdn_out_norm_f32_per_row","kernels":["norm::gated_rms_norm_f32_input"],"where":"device","numerics":"reference","compute":"memory","nodes":[431]},{"id":261,"rule":"w8a16_m32","kernels":["w8a16_gemm_pipelined_m32::w8a16_gemm_pipelined_m32"],"where":"device","numerics":"reference","compute":"tensor_core:mma.sync.m16n8k16.bf16","nodes":[432]},{"id":262,"rule":"residual_add_post_norm","kernels":["norm::residual_add_rms_norm"],"where":"device","numerics":"reference","compute":"memory","nodes":[433,434]},{"id":263,"rule":"moe_router_gemv","kernels":["gemv::dense_gemv_bf16"],"where":"device","numerics":"reference","compute":"cuda_core","nodes":[435]},{"id":264,"rule":"moe_topk_softmax","kernels":["moe_topk::moe_topk_softmax"],"where":"device","numerics":"reference","compute":"memory","nodes":[436]},{"id":265,"rule":"moe_gate_up_shared_fp8","kernels":["moe_shared_expert_fused_fp8::moe_expert_gate_up_shared_fp8"],"where":"device","numerics":"reference","compute":"cuda_core","nodes":[437,438]},{"id":266,"rule":"moe_silu_down_shared_fp8","kernels":["moe_shared_expert_fused_fp8::moe_expert_silu_down_shared_fp8"],"where":"device","numerics":"reference","compute":"cuda_core","nodes":[439,440,441,442]},{"id":267,"rule":"moe_weighted_sum_blend","kernels":["moe_expert_gemv::moe_weighted_sum_blend"],"where":"device","numerics":"reference","compute":"cuda_core","nodes":[443,444]},{"id":268,"rule":"cross_layer_add_norm","kernels":["residual_add_rms_norm_exact::residual_add_rms_norm_exact"],"where":"device","numerics":"bit_identical(residual_add_rms_norm_exact_microtest)","compute":"memory","nodes":[445,446]},{"id":269,"rule":"w8a16_m32","kernels":["w8a16_gemm_pipelined_m32::w8a16_gemm_pipelined_m32"],"where":"device","numerics":"reference","compute":"tensor_core:mma.sync.m16n8k16.bf16","nodes":[447]},{"id":270,"rule":"deinterleave_qg","kernels":["ssm_preprocess::deinterleave_qg"],"where":"device","numerics":"reference","compute":"memory","nodes":[448]},{"id":271,"rule":"w8a16_m32","kernels":["w8a16_gemm_pipelined_m32::w8a16_gemm_pipelined_m32"],"where":"device","numerics":"reference","compute":"tensor_core:mma.sync.m16n8k16.bf16","nodes":[449]},{"id":272,"rule":"w8a16_m32","kernels":["w8a16_gemm_pipelined_m32::w8a16_gemm_pipelined_m32"],"where":"device","numerics":"reference","compute":"tensor_core:mma.sync.m16n8k16.bf16","nodes":[450]},{"id":273,"rule":"qk_norm_rows","kernels":["norm::rms_norm"],"where":"device","numerics":"reference","compute":"memory","nodes":[451]},{"id":274,"rule":"qk_norm_rows","kernels":["norm::rms_norm"],"where":"device","numerics":"reference","compute":"memory","nodes":[452]},{"id":275,"rule":"rope_mrope_interleaved","kernels":["rope_mrope_interleaved::rope_forward_mrope_interleaved"],"where":"device","numerics":"reference","compute":"memory","nodes":[453]},{"id":276,"rule":"kv_write_bf16","kernels":["reshape_and_cache::reshape_and_cache_flash"],"where":"device","numerics":"reference","compute":"memory","nodes":[454]},{"id":277,"rule":"paged_attention_bf16","kernels":["paged_decode::paged_decode_attn"],"where":"device","numerics":"reference","compute":"cuda_core","nodes":[455]},{"id":278,"rule":"sigmoid_gate_mul","kernels":["residual_add::sigmoid_gate_mul"],"where":"device","numerics":"reference","compute":"memory","nodes":[456]},{"id":279,"rule":"w8a16_m32","kernels":["w8a16_gemm_pipelined_m32::w8a16_gemm_pipelined_m32"],"where":"device","numerics":"reference","compute":"tensor_core:mma.sync.m16n8k16.bf16","nodes":[457]},{"id":280,"rule":"residual_add_post_norm","kernels":["norm::residual_add_rms_norm"],"where":"device","numerics":"reference","compute":"memory","nodes":[458,459]},{"id":281,"rule":"moe_router_gemv","kernels":["gemv::dense_gemv_bf16"],"where":"device","numerics":"reference","compute":"cuda_core","nodes":[460]},{"id":282,"rule":"moe_topk_softmax","kernels":["moe_topk::moe_topk_softmax"],"where":"device","numerics":"reference","compute":"memory","nodes":[461]},{"id":283,"rule":"moe_gate_up_shared_fp8","kernels":["moe_shared_expert_fused_fp8::moe_expert_gate_up_shared_fp8"],"where":"device","numerics":"reference","compute":"cuda_core","nodes":[462,463]},{"id":284,"rule":"moe_silu_down_shared_fp8","kernels":["moe_shared_expert_fused_fp8::moe_expert_silu_down_shared_fp8"],"where":"device","numerics":"reference","compute":"cuda_core","nodes":[464,465,466,467]},{"id":285,"rule":"moe_weighted_sum_blend","kernels":["moe_expert_gemv::moe_weighted_sum_blend"],"where":"device","numerics":"reference","compute":"cuda_core","nodes":[468,469]},{"id":286,"rule":"cross_layer_add_norm","kernels":["residual_add_rms_norm_exact::residual_add_rms_norm_exact"],"where":"device","numerics":"bit_identical(residual_add_rms_norm_exact_microtest)","compute":"memory","nodes":[470,471]},{"id":287,"rule":"w8a16_m32","kernels":["w8a16_gemm_pipelined_m32::w8a16_gemm_pipelined_m32"],"where":"device","numerics":"reference","compute":"tensor_core:mma.sync.m16n8k16.bf16","nodes":[472]},{"id":288,"rule":"gdn_ba_gates_gemv_per_row","kernels":["ssm_preprocess::dense_gemv_ba_gates"],"where":"device","numerics":"reference","compute":"cuda_core","nodes":[473,474]},{"id":289,"rule":"gdn_conv_l2_f32_per_row","kernels":["causal_conv1d::causal_conv1d_update_l2norm_f32"],"where":"device","numerics":"reference","compute":"cuda_core","nodes":[475,476,477]},{"id":290,"rule":"gdn_recurrence_f32_per_row","kernels":["gated_delta_rule::gated_delta_rule_decode_f32"],"where":"device","numerics":"reference","compute":"cuda_core","nodes":[478]},{"id":291,"rule":"gdn_out_norm_f32_per_row","kernels":["norm::gated_rms_norm_f32_input"],"where":"device","numerics":"reference","compute":"memory","nodes":[479]},{"id":292,"rule":"w8a16_m32","kernels":["w8a16_gemm_pipelined_m32::w8a16_gemm_pipelined_m32"],"where":"device","numerics":"reference","compute":"tensor_core:mma.sync.m16n8k16.bf16","nodes":[480]},{"id":293,"rule":"residual_add_post_norm","kernels":["norm::residual_add_rms_norm"],"where":"device","numerics":"reference","compute":"memory","nodes":[481,482]},{"id":294,"rule":"moe_router_gemv","kernels":["gemv::dense_gemv_bf16"],"where":"device","numerics":"reference","compute":"cuda_core","nodes":[483]},{"id":295,"rule":"moe_topk_softmax","kernels":["moe_topk::moe_topk_softmax"],"where":"device","numerics":"reference","compute":"memory","nodes":[484]},{"id":296,"rule":"moe_gate_up_shared_fp8","kernels":["moe_shared_expert_fused_fp8::moe_expert_gate_up_shared_fp8"],"where":"device","numerics":"reference","compute":"cuda_core","nodes":[485,486]},{"id":297,"rule":"moe_silu_down_shared_fp8","kernels":["moe_shared_expert_fused_fp8::moe_expert_silu_down_shared_fp8"],"where":"device","numerics":"reference","compute":"cuda_core","nodes":[487,488,489,490]},{"id":298,"rule":"moe_weighted_sum_blend","kernels":["moe_expert_gemv::moe_weighted_sum_blend"],"where":"device","numerics":"reference","compute":"cuda_core","nodes":[491,492]},{"id":299,"rule":"cross_layer_add_norm","kernels":["residual_add_rms_norm_exact::residual_add_rms_norm_exact"],"where":"device","numerics":"bit_identical(residual_add_rms_norm_exact_microtest)","compute":"memory","nodes":[493,494]},{"id":300,"rule":"w8a16_m32","kernels":["w8a16_gemm_pipelined_m32::w8a16_gemm_pipelined_m32"],"where":"device","numerics":"reference","compute":"tensor_core:mma.sync.m16n8k16.bf16","nodes":[495]},{"id":301,"rule":"gdn_ba_gates_gemv_per_row","kernels":["ssm_preprocess::dense_gemv_ba_gates"],"where":"device","numerics":"reference","compute":"cuda_core","nodes":[496,497]},{"id":302,"rule":"gdn_conv_l2_f32_per_row","kernels":["causal_conv1d::causal_conv1d_update_l2norm_f32"],"where":"device","numerics":"reference","compute":"cuda_core","nodes":[498,499,500]},{"id":303,"rule":"gdn_recurrence_f32_per_row","kernels":["gated_delta_rule::gated_delta_rule_decode_f32"],"where":"device","numerics":"reference","compute":"cuda_core","nodes":[501]},{"id":304,"rule":"gdn_out_norm_f32_per_row","kernels":["norm::gated_rms_norm_f32_input"],"where":"device","numerics":"reference","compute":"memory","nodes":[502]},{"id":305,"rule":"w8a16_m32","kernels":["w8a16_gemm_pipelined_m32::w8a16_gemm_pipelined_m32"],"where":"device","numerics":"reference","compute":"tensor_core:mma.sync.m16n8k16.bf16","nodes":[503]},{"id":306,"rule":"residual_add_post_norm","kernels":["norm::residual_add_rms_norm"],"where":"device","numerics":"reference","compute":"memory","nodes":[504,505]},{"id":307,"rule":"moe_router_gemv","kernels":["gemv::dense_gemv_bf16"],"where":"device","numerics":"reference","compute":"cuda_core","nodes":[506]},{"id":308,"rule":"moe_topk_softmax","kernels":["moe_topk::moe_topk_softmax"],"where":"device","numerics":"reference","compute":"memory","nodes":[507]},{"id":309,"rule":"moe_gate_up_shared_fp8","kernels":["moe_shared_expert_fused_fp8::moe_expert_gate_up_shared_fp8"],"where":"device","numerics":"reference","compute":"cuda_core","nodes":[508,509]},{"id":310,"rule":"moe_silu_down_shared_fp8","kernels":["moe_shared_expert_fused_fp8::moe_expert_silu_down_shared_fp8"],"where":"device","numerics":"reference","compute":"cuda_core","nodes":[510,511,512,513]},{"id":311,"rule":"moe_weighted_sum_blend","kernels":["moe_expert_gemv::moe_weighted_sum_blend"],"where":"device","numerics":"reference","compute":"cuda_core","nodes":[514,515]},{"id":312,"rule":"cross_layer_add_norm","kernels":["residual_add_rms_norm_exact::residual_add_rms_norm_exact"],"where":"device","numerics":"bit_identical(residual_add_rms_norm_exact_microtest)","compute":"memory","nodes":[516,517]},{"id":313,"rule":"w8a16_m32","kernels":["w8a16_gemm_pipelined_m32::w8a16_gemm_pipelined_m32"],"where":"device","numerics":"reference","compute":"tensor_core:mma.sync.m16n8k16.bf16","nodes":[518]},{"id":314,"rule":"gdn_ba_gates_gemv_per_row","kernels":["ssm_preprocess::dense_gemv_ba_gates"],"where":"device","numerics":"reference","compute":"cuda_core","nodes":[519,520]},{"id":315,"rule":"gdn_conv_l2_f32_per_row","kernels":["causal_conv1d::causal_conv1d_update_l2norm_f32"],"where":"device","numerics":"reference","compute":"cuda_core","nodes":[521,522,523]},{"id":316,"rule":"gdn_recurrence_f32_per_row","kernels":["gated_delta_rule::gated_delta_rule_decode_f32"],"where":"device","numerics":"reference","compute":"cuda_core","nodes":[524]},{"id":317,"rule":"gdn_out_norm_f32_per_row","kernels":["norm::gated_rms_norm_f32_input"],"where":"device","numerics":"reference","compute":"memory","nodes":[525]},{"id":318,"rule":"w8a16_m32","kernels":["w8a16_gemm_pipelined_m32::w8a16_gemm_pipelined_m32"],"where":"device","numerics":"reference","compute":"tensor_core:mma.sync.m16n8k16.bf16","nodes":[526]},{"id":319,"rule":"residual_add_post_norm","kernels":["norm::residual_add_rms_norm"],"where":"device","numerics":"reference","compute":"memory","nodes":[527,528]},{"id":320,"rule":"moe_router_gemv","kernels":["gemv::dense_gemv_bf16"],"where":"device","numerics":"reference","compute":"cuda_core","nodes":[529]},{"id":321,"rule":"moe_topk_softmax","kernels":["moe_topk::moe_topk_softmax"],"where":"device","numerics":"reference","compute":"memory","nodes":[530]},{"id":322,"rule":"moe_gate_up_shared_fp8","kernels":["moe_shared_expert_fused_fp8::moe_expert_gate_up_shared_fp8"],"where":"device","numerics":"reference","compute":"cuda_core","nodes":[531,532]},{"id":323,"rule":"moe_silu_down_shared_fp8","kernels":["moe_shared_expert_fused_fp8::moe_expert_silu_down_shared_fp8"],"where":"device","numerics":"reference","compute":"cuda_core","nodes":[533,534,535,536]},{"id":324,"rule":"moe_weighted_sum_blend","kernels":["moe_expert_gemv::moe_weighted_sum_blend"],"where":"device","numerics":"reference","compute":"cuda_core","nodes":[537,538]},{"id":325,"rule":"cross_layer_add_norm","kernels":["residual_add_rms_norm_exact::residual_add_rms_norm_exact"],"where":"device","numerics":"bit_identical(residual_add_rms_norm_exact_microtest)","compute":"memory","nodes":[539,540]},{"id":326,"rule":"w8a16_m32","kernels":["w8a16_gemm_pipelined_m32::w8a16_gemm_pipelined_m32"],"where":"device","numerics":"reference","compute":"tensor_core:mma.sync.m16n8k16.bf16","nodes":[541]},{"id":327,"rule":"deinterleave_qg","kernels":["ssm_preprocess::deinterleave_qg"],"where":"device","numerics":"reference","compute":"memory","nodes":[542]},{"id":328,"rule":"w8a16_m32","kernels":["w8a16_gemm_pipelined_m32::w8a16_gemm_pipelined_m32"],"where":"device","numerics":"reference","compute":"tensor_core:mma.sync.m16n8k16.bf16","nodes":[543]},{"id":329,"rule":"w8a16_m32","kernels":["w8a16_gemm_pipelined_m32::w8a16_gemm_pipelined_m32"],"where":"device","numerics":"reference","compute":"tensor_core:mma.sync.m16n8k16.bf16","nodes":[544]},{"id":330,"rule":"qk_norm_rows","kernels":["norm::rms_norm"],"where":"device","numerics":"reference","compute":"memory","nodes":[545]},{"id":331,"rule":"qk_norm_rows","kernels":["norm::rms_norm"],"where":"device","numerics":"reference","compute":"memory","nodes":[546]},{"id":332,"rule":"rope_mrope_interleaved","kernels":["rope_mrope_interleaved::rope_forward_mrope_interleaved"],"where":"device","numerics":"reference","compute":"memory","nodes":[547]},{"id":333,"rule":"kv_write_bf16","kernels":["reshape_and_cache::reshape_and_cache_flash"],"where":"device","numerics":"reference","compute":"memory","nodes":[548]},{"id":334,"rule":"paged_attention_bf16","kernels":["paged_decode::paged_decode_attn"],"where":"device","numerics":"reference","compute":"cuda_core","nodes":[549]},{"id":335,"rule":"sigmoid_gate_mul","kernels":["residual_add::sigmoid_gate_mul"],"where":"device","numerics":"reference","compute":"memory","nodes":[550]},{"id":336,"rule":"w8a16_m32","kernels":["w8a16_gemm_pipelined_m32::w8a16_gemm_pipelined_m32"],"where":"device","numerics":"reference","compute":"tensor_core:mma.sync.m16n8k16.bf16","nodes":[551]},{"id":337,"rule":"residual_add_post_norm","kernels":["norm::residual_add_rms_norm"],"where":"device","numerics":"reference","compute":"memory","nodes":[552,553]},{"id":338,"rule":"moe_router_gemv","kernels":["gemv::dense_gemv_bf16"],"where":"device","numerics":"reference","compute":"cuda_core","nodes":[554]},{"id":339,"rule":"moe_topk_softmax","kernels":["moe_topk::moe_topk_softmax"],"where":"device","numerics":"reference","compute":"memory","nodes":[555]},{"id":340,"rule":"moe_gate_up_shared_fp8","kernels":["moe_shared_expert_fused_fp8::moe_expert_gate_up_shared_fp8"],"where":"device","numerics":"reference","compute":"cuda_core","nodes":[556,557]},{"id":341,"rule":"moe_silu_down_shared_fp8","kernels":["moe_shared_expert_fused_fp8::moe_expert_silu_down_shared_fp8"],"where":"device","numerics":"reference","compute":"cuda_core","nodes":[558,559,560,561]},{"id":342,"rule":"moe_weighted_sum_blend","kernels":["moe_expert_gemv::moe_weighted_sum_blend"],"where":"device","numerics":"reference","compute":"cuda_core","nodes":[562,563]},{"id":343,"rule":"cross_layer_add_norm","kernels":["residual_add_rms_norm_exact::residual_add_rms_norm_exact"],"where":"device","numerics":"bit_identical(residual_add_rms_norm_exact_microtest)","compute":"memory","nodes":[564,565]},{"id":344,"rule":"w8a16_m32","kernels":["w8a16_gemm_pipelined_m32::w8a16_gemm_pipelined_m32"],"where":"device","numerics":"reference","compute":"tensor_core:mma.sync.m16n8k16.bf16","nodes":[566]},{"id":345,"rule":"gdn_ba_gates_gemv_per_row","kernels":["ssm_preprocess::dense_gemv_ba_gates"],"where":"device","numerics":"reference","compute":"cuda_core","nodes":[567,568]},{"id":346,"rule":"gdn_conv_l2_f32_per_row","kernels":["causal_conv1d::causal_conv1d_update_l2norm_f32"],"where":"device","numerics":"reference","compute":"cuda_core","nodes":[569,570,571]},{"id":347,"rule":"gdn_recurrence_f32_per_row","kernels":["gated_delta_rule::gated_delta_rule_decode_f32"],"where":"device","numerics":"reference","compute":"cuda_core","nodes":[572]},{"id":348,"rule":"gdn_out_norm_f32_per_row","kernels":["norm::gated_rms_norm_f32_input"],"where":"device","numerics":"reference","compute":"memory","nodes":[573]},{"id":349,"rule":"w8a16_m32","kernels":["w8a16_gemm_pipelined_m32::w8a16_gemm_pipelined_m32"],"where":"device","numerics":"reference","compute":"tensor_core:mma.sync.m16n8k16.bf16","nodes":[574]},{"id":350,"rule":"residual_add_post_norm","kernels":["norm::residual_add_rms_norm"],"where":"device","numerics":"reference","compute":"memory","nodes":[575,576]},{"id":351,"rule":"moe_router_gemv","kernels":["gemv::dense_gemv_bf16"],"where":"device","numerics":"reference","compute":"cuda_core","nodes":[577]},{"id":352,"rule":"moe_topk_softmax","kernels":["moe_topk::moe_topk_softmax"],"where":"device","numerics":"reference","compute":"memory","nodes":[578]},{"id":353,"rule":"moe_gate_up_shared_fp8","kernels":["moe_shared_expert_fused_fp8::moe_expert_gate_up_shared_fp8"],"where":"device","numerics":"reference","compute":"cuda_core","nodes":[579,580]},{"id":354,"rule":"moe_silu_down_shared_fp8","kernels":["moe_shared_expert_fused_fp8::moe_expert_silu_down_shared_fp8"],"where":"device","numerics":"reference","compute":"cuda_core","nodes":[581,582,583,584]},{"id":355,"rule":"moe_weighted_sum_blend","kernels":["moe_expert_gemv::moe_weighted_sum_blend"],"where":"device","numerics":"reference","compute":"cuda_core","nodes":[585,586]},{"id":356,"rule":"cross_layer_add_norm","kernels":["residual_add_rms_norm_exact::residual_add_rms_norm_exact"],"where":"device","numerics":"bit_identical(residual_add_rms_norm_exact_microtest)","compute":"memory","nodes":[587,588]},{"id":357,"rule":"w8a16_m32","kernels":["w8a16_gemm_pipelined_m32::w8a16_gemm_pipelined_m32"],"where":"device","numerics":"reference","compute":"tensor_core:mma.sync.m16n8k16.bf16","nodes":[589]},{"id":358,"rule":"gdn_ba_gates_gemv_per_row","kernels":["ssm_preprocess::dense_gemv_ba_gates"],"where":"device","numerics":"reference","compute":"cuda_core","nodes":[590,591]},{"id":359,"rule":"gdn_conv_l2_f32_per_row","kernels":["causal_conv1d::causal_conv1d_update_l2norm_f32"],"where":"device","numerics":"reference","compute":"cuda_core","nodes":[592,593,594]},{"id":360,"rule":"gdn_recurrence_f32_per_row","kernels":["gated_delta_rule::gated_delta_rule_decode_f32"],"where":"device","numerics":"reference","compute":"cuda_core","nodes":[595]},{"id":361,"rule":"gdn_out_norm_f32_per_row","kernels":["norm::gated_rms_norm_f32_input"],"where":"device","numerics":"reference","compute":"memory","nodes":[596]},{"id":362,"rule":"w8a16_m32","kernels":["w8a16_gemm_pipelined_m32::w8a16_gemm_pipelined_m32"],"where":"device","numerics":"reference","compute":"tensor_core:mma.sync.m16n8k16.bf16","nodes":[597]},{"id":363,"rule":"residual_add_post_norm","kernels":["norm::residual_add_rms_norm"],"where":"device","numerics":"reference","compute":"memory","nodes":[598,599]},{"id":364,"rule":"moe_router_gemv","kernels":["gemv::dense_gemv_bf16"],"where":"device","numerics":"reference","compute":"cuda_core","nodes":[600]},{"id":365,"rule":"moe_topk_softmax","kernels":["moe_topk::moe_topk_softmax"],"where":"device","numerics":"reference","compute":"memory","nodes":[601]},{"id":366,"rule":"moe_gate_up_shared_fp8","kernels":["moe_shared_expert_fused_fp8::moe_expert_gate_up_shared_fp8"],"where":"device","numerics":"reference","compute":"cuda_core","nodes":[602,603]},{"id":367,"rule":"moe_silu_down_shared_fp8","kernels":["moe_shared_expert_fused_fp8::moe_expert_silu_down_shared_fp8"],"where":"device","numerics":"reference","compute":"cuda_core","nodes":[604,605,606,607]},{"id":368,"rule":"moe_weighted_sum_blend","kernels":["moe_expert_gemv::moe_weighted_sum_blend"],"where":"device","numerics":"reference","compute":"cuda_core","nodes":[608,609]},{"id":369,"rule":"cross_layer_add_norm","kernels":["residual_add_rms_norm_exact::residual_add_rms_norm_exact"],"where":"device","numerics":"bit_identical(residual_add_rms_norm_exact_microtest)","compute":"memory","nodes":[610,611]},{"id":370,"rule":"w8a16_m32","kernels":["w8a16_gemm_pipelined_m32::w8a16_gemm_pipelined_m32"],"where":"device","numerics":"reference","compute":"tensor_core:mma.sync.m16n8k16.bf16","nodes":[612]},{"id":371,"rule":"gdn_ba_gates_gemv_per_row","kernels":["ssm_preprocess::dense_gemv_ba_gates"],"where":"device","numerics":"reference","compute":"cuda_core","nodes":[613,614]},{"id":372,"rule":"gdn_conv_l2_f32_per_row","kernels":["causal_conv1d::causal_conv1d_update_l2norm_f32"],"where":"device","numerics":"reference","compute":"cuda_core","nodes":[615,616,617]},{"id":373,"rule":"gdn_recurrence_f32_per_row","kernels":["gated_delta_rule::gated_delta_rule_decode_f32"],"where":"device","numerics":"reference","compute":"cuda_core","nodes":[618]},{"id":374,"rule":"gdn_out_norm_f32_per_row","kernels":["norm::gated_rms_norm_f32_input"],"where":"device","numerics":"reference","compute":"memory","nodes":[619]},{"id":375,"rule":"w8a16_m32","kernels":["w8a16_gemm_pipelined_m32::w8a16_gemm_pipelined_m32"],"where":"device","numerics":"reference","compute":"tensor_core:mma.sync.m16n8k16.bf16","nodes":[620]},{"id":376,"rule":"residual_add_post_norm","kernels":["norm::residual_add_rms_norm"],"where":"device","numerics":"reference","compute":"memory","nodes":[621,622]},{"id":377,"rule":"moe_router_gemv","kernels":["gemv::dense_gemv_bf16"],"where":"device","numerics":"reference","compute":"cuda_core","nodes":[623]},{"id":378,"rule":"moe_topk_softmax","kernels":["moe_topk::moe_topk_softmax"],"where":"device","numerics":"reference","compute":"memory","nodes":[624]},{"id":379,"rule":"moe_gate_up_shared_fp8","kernels":["moe_shared_expert_fused_fp8::moe_expert_gate_up_shared_fp8"],"where":"device","numerics":"reference","compute":"cuda_core","nodes":[625,626]},{"id":380,"rule":"moe_silu_down_shared_fp8","kernels":["moe_shared_expert_fused_fp8::moe_expert_silu_down_shared_fp8"],"where":"device","numerics":"reference","compute":"cuda_core","nodes":[627,628,629,630]},{"id":381,"rule":"moe_weighted_sum_blend","kernels":["moe_expert_gemv::moe_weighted_sum_blend"],"where":"device","numerics":"reference","compute":"cuda_core","nodes":[631,632]},{"id":382,"rule":"cross_layer_add_norm","kernels":["residual_add_rms_norm_exact::residual_add_rms_norm_exact"],"where":"device","numerics":"bit_identical(residual_add_rms_norm_exact_microtest)","compute":"memory","nodes":[633,634]},{"id":383,"rule":"w8a16_m32","kernels":["w8a16_gemm_pipelined_m32::w8a16_gemm_pipelined_m32"],"where":"device","numerics":"reference","compute":"tensor_core:mma.sync.m16n8k16.bf16","nodes":[635]},{"id":384,"rule":"deinterleave_qg","kernels":["ssm_preprocess::deinterleave_qg"],"where":"device","numerics":"reference","compute":"memory","nodes":[636]},{"id":385,"rule":"w8a16_m32","kernels":["w8a16_gemm_pipelined_m32::w8a16_gemm_pipelined_m32"],"where":"device","numerics":"reference","compute":"tensor_core:mma.sync.m16n8k16.bf16","nodes":[637]},{"id":386,"rule":"w8a16_m32","kernels":["w8a16_gemm_pipelined_m32::w8a16_gemm_pipelined_m32"],"where":"device","numerics":"reference","compute":"tensor_core:mma.sync.m16n8k16.bf16","nodes":[638]},{"id":387,"rule":"qk_norm_rows","kernels":["norm::rms_norm"],"where":"device","numerics":"reference","compute":"memory","nodes":[639]},{"id":388,"rule":"qk_norm_rows","kernels":["norm::rms_norm"],"where":"device","numerics":"reference","compute":"memory","nodes":[640]},{"id":389,"rule":"rope_mrope_interleaved","kernels":["rope_mrope_interleaved::rope_forward_mrope_interleaved"],"where":"device","numerics":"reference","compute":"memory","nodes":[641]},{"id":390,"rule":"kv_write_bf16","kernels":["reshape_and_cache::reshape_and_cache_flash"],"where":"device","numerics":"reference","compute":"memory","nodes":[642]},{"id":391,"rule":"paged_attention_bf16","kernels":["paged_decode::paged_decode_attn"],"where":"device","numerics":"reference","compute":"cuda_core","nodes":[643]},{"id":392,"rule":"sigmoid_gate_mul","kernels":["residual_add::sigmoid_gate_mul"],"where":"device","numerics":"reference","compute":"memory","nodes":[644]},{"id":393,"rule":"w8a16_m32","kernels":["w8a16_gemm_pipelined_m32::w8a16_gemm_pipelined_m32"],"where":"device","numerics":"reference","compute":"tensor_core:mma.sync.m16n8k16.bf16","nodes":[645]},{"id":394,"rule":"residual_add_post_norm","kernels":["norm::residual_add_rms_norm"],"where":"device","numerics":"reference","compute":"memory","nodes":[646,647]},{"id":395,"rule":"moe_router_gemv","kernels":["gemv::dense_gemv_bf16"],"where":"device","numerics":"reference","compute":"cuda_core","nodes":[648]},{"id":396,"rule":"moe_topk_softmax","kernels":["moe_topk::moe_topk_softmax"],"where":"device","numerics":"reference","compute":"memory","nodes":[649]},{"id":397,"rule":"moe_gate_up_shared_fp8","kernels":["moe_shared_expert_fused_fp8::moe_expert_gate_up_shared_fp8"],"where":"device","numerics":"reference","compute":"cuda_core","nodes":[650,651]},{"id":398,"rule":"moe_silu_down_shared_fp8","kernels":["moe_shared_expert_fused_fp8::moe_expert_silu_down_shared_fp8"],"where":"device","numerics":"reference","compute":"cuda_core","nodes":[652,653,654,655]},{"id":399,"rule":"moe_weighted_sum_blend","kernels":["moe_expert_gemv::moe_weighted_sum_blend"],"where":"device","numerics":"reference","compute":"cuda_core","nodes":[656,657]},{"id":400,"rule":"cross_layer_add_norm","kernels":["residual_add_rms_norm_exact::residual_add_rms_norm_exact"],"where":"device","numerics":"bit_identical(residual_add_rms_norm_exact_microtest)","compute":"memory","nodes":[658,659]},{"id":401,"rule":"w8a16_m32","kernels":["w8a16_gemm_pipelined_m32::w8a16_gemm_pipelined_m32"],"where":"device","numerics":"reference","compute":"tensor_core:mma.sync.m16n8k16.bf16","nodes":[660]},{"id":402,"rule":"gdn_ba_gates_gemv_per_row","kernels":["ssm_preprocess::dense_gemv_ba_gates"],"where":"device","numerics":"reference","compute":"cuda_core","nodes":[661,662]},{"id":403,"rule":"gdn_conv_l2_f32_per_row","kernels":["causal_conv1d::causal_conv1d_update_l2norm_f32"],"where":"device","numerics":"reference","compute":"cuda_core","nodes":[663,664,665]},{"id":404,"rule":"gdn_recurrence_f32_per_row","kernels":["gated_delta_rule::gated_delta_rule_decode_f32"],"where":"device","numerics":"reference","compute":"cuda_core","nodes":[666]},{"id":405,"rule":"gdn_out_norm_f32_per_row","kernels":["norm::gated_rms_norm_f32_input"],"where":"device","numerics":"reference","compute":"memory","nodes":[667]},{"id":406,"rule":"w8a16_m32","kernels":["w8a16_gemm_pipelined_m32::w8a16_gemm_pipelined_m32"],"where":"device","numerics":"reference","compute":"tensor_core:mma.sync.m16n8k16.bf16","nodes":[668]},{"id":407,"rule":"residual_add_post_norm","kernels":["norm::residual_add_rms_norm"],"where":"device","numerics":"reference","compute":"memory","nodes":[669,670]},{"id":408,"rule":"moe_router_gemv","kernels":["gemv::dense_gemv_bf16"],"where":"device","numerics":"reference","compute":"cuda_core","nodes":[671]},{"id":409,"rule":"moe_topk_softmax","kernels":["moe_topk::moe_topk_softmax"],"where":"device","numerics":"reference","compute":"memory","nodes":[672]},{"id":410,"rule":"moe_gate_up_shared_fp8","kernels":["moe_shared_expert_fused_fp8::moe_expert_gate_up_shared_fp8"],"where":"device","numerics":"reference","compute":"cuda_core","nodes":[673,674]},{"id":411,"rule":"moe_silu_down_shared_fp8","kernels":["moe_shared_expert_fused_fp8::moe_expert_silu_down_shared_fp8"],"where":"device","numerics":"reference","compute":"cuda_core","nodes":[675,676,677,678]},{"id":412,"rule":"moe_weighted_sum_blend","kernels":["moe_expert_gemv::moe_weighted_sum_blend"],"where":"device","numerics":"reference","compute":"cuda_core","nodes":[679,680]},{"id":413,"rule":"cross_layer_add_norm","kernels":["residual_add_rms_norm_exact::residual_add_rms_norm_exact"],"where":"device","numerics":"bit_identical(residual_add_rms_norm_exact_microtest)","compute":"memory","nodes":[681,682]},{"id":414,"rule":"w8a16_m32","kernels":["w8a16_gemm_pipelined_m32::w8a16_gemm_pipelined_m32"],"where":"device","numerics":"reference","compute":"tensor_core:mma.sync.m16n8k16.bf16","nodes":[683]},{"id":415,"rule":"gdn_ba_gates_gemv_per_row","kernels":["ssm_preprocess::dense_gemv_ba_gates"],"where":"device","numerics":"reference","compute":"cuda_core","nodes":[684,685]},{"id":416,"rule":"gdn_conv_l2_f32_per_row","kernels":["causal_conv1d::causal_conv1d_update_l2norm_f32"],"where":"device","numerics":"reference","compute":"cuda_core","nodes":[686,687,688]},{"id":417,"rule":"gdn_recurrence_f32_per_row","kernels":["gated_delta_rule::gated_delta_rule_decode_f32"],"where":"device","numerics":"reference","compute":"cuda_core","nodes":[689]},{"id":418,"rule":"gdn_out_norm_f32_per_row","kernels":["norm::gated_rms_norm_f32_input"],"where":"device","numerics":"reference","compute":"memory","nodes":[690]},{"id":419,"rule":"w8a16_m32","kernels":["w8a16_gemm_pipelined_m32::w8a16_gemm_pipelined_m32"],"where":"device","numerics":"reference","compute":"tensor_core:mma.sync.m16n8k16.bf16","nodes":[691]},{"id":420,"rule":"residual_add_post_norm","kernels":["norm::residual_add_rms_norm"],"where":"device","numerics":"reference","compute":"memory","nodes":[692,693]},{"id":421,"rule":"moe_router_gemv","kernels":["gemv::dense_gemv_bf16"],"where":"device","numerics":"reference","compute":"cuda_core","nodes":[694]},{"id":422,"rule":"moe_topk_softmax","kernels":["moe_topk::moe_topk_softmax"],"where":"device","numerics":"reference","compute":"memory","nodes":[695]},{"id":423,"rule":"moe_gate_up_shared_fp8","kernels":["moe_shared_expert_fused_fp8::moe_expert_gate_up_shared_fp8"],"where":"device","numerics":"reference","compute":"cuda_core","nodes":[696,697]},{"id":424,"rule":"moe_silu_down_shared_fp8","kernels":["moe_shared_expert_fused_fp8::moe_expert_silu_down_shared_fp8"],"where":"device","numerics":"reference","compute":"cuda_core","nodes":[698,699,700,701]},{"id":425,"rule":"moe_weighted_sum_blend","kernels":["moe_expert_gemv::moe_weighted_sum_blend"],"where":"device","numerics":"reference","compute":"cuda_core","nodes":[702,703]},{"id":426,"rule":"cross_layer_add_norm","kernels":["residual_add_rms_norm_exact::residual_add_rms_norm_exact"],"where":"device","numerics":"bit_identical(residual_add_rms_norm_exact_microtest)","compute":"memory","nodes":[704,705]},{"id":427,"rule":"w8a16_m32","kernels":["w8a16_gemm_pipelined_m32::w8a16_gemm_pipelined_m32"],"where":"device","numerics":"reference","compute":"tensor_core:mma.sync.m16n8k16.bf16","nodes":[706]},{"id":428,"rule":"gdn_ba_gates_gemv_per_row","kernels":["ssm_preprocess::dense_gemv_ba_gates"],"where":"device","numerics":"reference","compute":"cuda_core","nodes":[707,708]},{"id":429,"rule":"gdn_conv_l2_f32_per_row","kernels":["causal_conv1d::causal_conv1d_update_l2norm_f32"],"where":"device","numerics":"reference","compute":"cuda_core","nodes":[709,710,711]},{"id":430,"rule":"gdn_recurrence_f32_per_row","kernels":["gated_delta_rule::gated_delta_rule_decode_f32"],"where":"device","numerics":"reference","compute":"cuda_core","nodes":[712]},{"id":431,"rule":"gdn_out_norm_f32_per_row","kernels":["norm::gated_rms_norm_f32_input"],"where":"device","numerics":"reference","compute":"memory","nodes":[713]},{"id":432,"rule":"w8a16_m32","kernels":["w8a16_gemm_pipelined_m32::w8a16_gemm_pipelined_m32"],"where":"device","numerics":"reference","compute":"tensor_core:mma.sync.m16n8k16.bf16","nodes":[714]},{"id":433,"rule":"residual_add_post_norm","kernels":["norm::residual_add_rms_norm"],"where":"device","numerics":"reference","compute":"memory","nodes":[715,716]},{"id":434,"rule":"moe_router_gemv","kernels":["gemv::dense_gemv_bf16"],"where":"device","numerics":"reference","compute":"cuda_core","nodes":[717]},{"id":435,"rule":"moe_topk_softmax","kernels":["moe_topk::moe_topk_softmax"],"where":"device","numerics":"reference","compute":"memory","nodes":[718]},{"id":436,"rule":"moe_gate_up_shared_fp8","kernels":["moe_shared_expert_fused_fp8::moe_expert_gate_up_shared_fp8"],"where":"device","numerics":"reference","compute":"cuda_core","nodes":[719,720]},{"id":437,"rule":"moe_silu_down_shared_fp8","kernels":["moe_shared_expert_fused_fp8::moe_expert_silu_down_shared_fp8"],"where":"device","numerics":"reference","compute":"cuda_core","nodes":[721,722,723,724]},{"id":438,"rule":"moe_weighted_sum_blend","kernels":["moe_expert_gemv::moe_weighted_sum_blend"],"where":"device","numerics":"reference","compute":"cuda_core","nodes":[725,726]},{"id":439,"rule":"cross_layer_add_norm","kernels":["residual_add_rms_norm_exact::residual_add_rms_norm_exact"],"where":"device","numerics":"bit_identical(residual_add_rms_norm_exact_microtest)","compute":"memory","nodes":[727,728]},{"id":440,"rule":"w8a16_m32","kernels":["w8a16_gemm_pipelined_m32::w8a16_gemm_pipelined_m32"],"where":"device","numerics":"reference","compute":"tensor_core:mma.sync.m16n8k16.bf16","nodes":[729]},{"id":441,"rule":"deinterleave_qg","kernels":["ssm_preprocess::deinterleave_qg"],"where":"device","numerics":"reference","compute":"memory","nodes":[730]},{"id":442,"rule":"w8a16_m32","kernels":["w8a16_gemm_pipelined_m32::w8a16_gemm_pipelined_m32"],"where":"device","numerics":"reference","compute":"tensor_core:mma.sync.m16n8k16.bf16","nodes":[731]},{"id":443,"rule":"w8a16_m32","kernels":["w8a16_gemm_pipelined_m32::w8a16_gemm_pipelined_m32"],"where":"device","numerics":"reference","compute":"tensor_core:mma.sync.m16n8k16.bf16","nodes":[732]},{"id":444,"rule":"qk_norm_rows","kernels":["norm::rms_norm"],"where":"device","numerics":"reference","compute":"memory","nodes":[733]},{"id":445,"rule":"qk_norm_rows","kernels":["norm::rms_norm"],"where":"device","numerics":"reference","compute":"memory","nodes":[734]},{"id":446,"rule":"rope_mrope_interleaved","kernels":["rope_mrope_interleaved::rope_forward_mrope_interleaved"],"where":"device","numerics":"reference","compute":"memory","nodes":[735]},{"id":447,"rule":"kv_write_bf16","kernels":["reshape_and_cache::reshape_and_cache_flash"],"where":"device","numerics":"reference","compute":"memory","nodes":[736]},{"id":448,"rule":"paged_attention_bf16","kernels":["paged_decode::paged_decode_attn"],"where":"device","numerics":"reference","compute":"cuda_core","nodes":[737]},{"id":449,"rule":"sigmoid_gate_mul","kernels":["residual_add::sigmoid_gate_mul"],"where":"device","numerics":"reference","compute":"memory","nodes":[738]},{"id":450,"rule":"w8a16_m32","kernels":["w8a16_gemm_pipelined_m32::w8a16_gemm_pipelined_m32"],"where":"device","numerics":"reference","compute":"tensor_core:mma.sync.m16n8k16.bf16","nodes":[739]},{"id":451,"rule":"residual_add_post_norm","kernels":["norm::residual_add_rms_norm"],"where":"device","numerics":"reference","compute":"memory","nodes":[740,741]},{"id":452,"rule":"moe_router_gemv","kernels":["gemv::dense_gemv_bf16"],"where":"device","numerics":"reference","compute":"cuda_core","nodes":[742]},{"id":453,"rule":"moe_topk_softmax","kernels":["moe_topk::moe_topk_softmax"],"where":"device","numerics":"reference","compute":"memory","nodes":[743]},{"id":454,"rule":"moe_gate_up_shared_fp8","kernels":["moe_shared_expert_fused_fp8::moe_expert_gate_up_shared_fp8"],"where":"device","numerics":"reference","compute":"cuda_core","nodes":[744,745]},{"id":455,"rule":"moe_silu_down_shared_fp8","kernels":["moe_shared_expert_fused_fp8::moe_expert_silu_down_shared_fp8"],"where":"device","numerics":"reference","compute":"cuda_core","nodes":[746,747,748,749]},{"id":456,"rule":"moe_weighted_sum_blend","kernels":["moe_expert_gemv::moe_weighted_sum_blend"],"where":"device","numerics":"reference","compute":"cuda_core","nodes":[750,751]},{"id":457,"rule":"cross_layer_add_norm","kernels":["residual_add_rms_norm_exact::residual_add_rms_norm_exact"],"where":"device","numerics":"bit_identical(residual_add_rms_norm_exact_microtest)","compute":"memory","nodes":[752,753]},{"id":458,"rule":"w8a16_m32","kernels":["w8a16_gemm_pipelined_m32::w8a16_gemm_pipelined_m32"],"where":"device","numerics":"reference","compute":"tensor_core:mma.sync.m16n8k16.bf16","nodes":[754]},{"id":459,"rule":"gdn_ba_gates_gemv_per_row","kernels":["ssm_preprocess::dense_gemv_ba_gates"],"where":"device","numerics":"reference","compute":"cuda_core","nodes":[755,756]},{"id":460,"rule":"gdn_conv_l2_f32_per_row","kernels":["causal_conv1d::causal_conv1d_update_l2norm_f32"],"where":"device","numerics":"reference","compute":"cuda_core","nodes":[757,758,759]},{"id":461,"rule":"gdn_recurrence_f32_per_row","kernels":["gated_delta_rule::gated_delta_rule_decode_f32"],"where":"device","numerics":"reference","compute":"cuda_core","nodes":[760]},{"id":462,"rule":"gdn_out_norm_f32_per_row","kernels":["norm::gated_rms_norm_f32_input"],"where":"device","numerics":"reference","compute":"memory","nodes":[761]},{"id":463,"rule":"w8a16_m32","kernels":["w8a16_gemm_pipelined_m32::w8a16_gemm_pipelined_m32"],"where":"device","numerics":"reference","compute":"tensor_core:mma.sync.m16n8k16.bf16","nodes":[762]},{"id":464,"rule":"residual_add_post_norm","kernels":["norm::residual_add_rms_norm"],"where":"device","numerics":"reference","compute":"memory","nodes":[763,764]},{"id":465,"rule":"moe_router_gemv","kernels":["gemv::dense_gemv_bf16"],"where":"device","numerics":"reference","compute":"cuda_core","nodes":[765]},{"id":466,"rule":"moe_topk_softmax","kernels":["moe_topk::moe_topk_softmax"],"where":"device","numerics":"reference","compute":"memory","nodes":[766]},{"id":467,"rule":"moe_gate_up_shared_fp8","kernels":["moe_shared_expert_fused_fp8::moe_expert_gate_up_shared_fp8"],"where":"device","numerics":"reference","compute":"cuda_core","nodes":[767,768]},{"id":468,"rule":"moe_silu_down_shared_fp8","kernels":["moe_shared_expert_fused_fp8::moe_expert_silu_down_shared_fp8"],"where":"device","numerics":"reference","compute":"cuda_core","nodes":[769,770,771,772]},{"id":469,"rule":"moe_weighted_sum_blend","kernels":["moe_expert_gemv::moe_weighted_sum_blend"],"where":"device","numerics":"reference","compute":"cuda_core","nodes":[773,774]},{"id":470,"rule":"cross_layer_add_norm","kernels":["residual_add_rms_norm_exact::residual_add_rms_norm_exact"],"where":"device","numerics":"bit_identical(residual_add_rms_norm_exact_microtest)","compute":"memory","nodes":[775,776]},{"id":471,"rule":"w8a16_m32","kernels":["w8a16_gemm_pipelined_m32::w8a16_gemm_pipelined_m32"],"where":"device","numerics":"reference","compute":"tensor_core:mma.sync.m16n8k16.bf16","nodes":[777]},{"id":472,"rule":"gdn_ba_gates_gemv_per_row","kernels":["ssm_preprocess::dense_gemv_ba_gates"],"where":"device","numerics":"reference","compute":"cuda_core","nodes":[778,779]},{"id":473,"rule":"gdn_conv_l2_f32_per_row","kernels":["causal_conv1d::causal_conv1d_update_l2norm_f32"],"where":"device","numerics":"reference","compute":"cuda_core","nodes":[780,781,782]},{"id":474,"rule":"gdn_recurrence_f32_per_row","kernels":["gated_delta_rule::gated_delta_rule_decode_f32"],"where":"device","numerics":"reference","compute":"cuda_core","nodes":[783]},{"id":475,"rule":"gdn_out_norm_f32_per_row","kernels":["norm::gated_rms_norm_f32_input"],"where":"device","numerics":"reference","compute":"memory","nodes":[784]},{"id":476,"rule":"w8a16_m32","kernels":["w8a16_gemm_pipelined_m32::w8a16_gemm_pipelined_m32"],"where":"device","numerics":"reference","compute":"tensor_core:mma.sync.m16n8k16.bf16","nodes":[785]},{"id":477,"rule":"residual_add_post_norm","kernels":["norm::residual_add_rms_norm"],"where":"device","numerics":"reference","compute":"memory","nodes":[786,787]},{"id":478,"rule":"moe_router_gemv","kernels":["gemv::dense_gemv_bf16"],"where":"device","numerics":"reference","compute":"cuda_core","nodes":[788]},{"id":479,"rule":"moe_topk_softmax","kernels":["moe_topk::moe_topk_softmax"],"where":"device","numerics":"reference","compute":"memory","nodes":[789]},{"id":480,"rule":"moe_gate_up_shared_fp8","kernels":["moe_shared_expert_fused_fp8::moe_expert_gate_up_shared_fp8"],"where":"device","numerics":"reference","compute":"cuda_core","nodes":[790,791]},{"id":481,"rule":"moe_silu_down_shared_fp8","kernels":["moe_shared_expert_fused_fp8::moe_expert_silu_down_shared_fp8"],"where":"device","numerics":"reference","compute":"cuda_core","nodes":[792,793,794,795]},{"id":482,"rule":"moe_weighted_sum_blend","kernels":["moe_expert_gemv::moe_weighted_sum_blend"],"where":"device","numerics":"reference","compute":"cuda_core","nodes":[796,797]},{"id":483,"rule":"cross_layer_add_norm","kernels":["residual_add_rms_norm_exact::residual_add_rms_norm_exact"],"where":"device","numerics":"bit_identical(residual_add_rms_norm_exact_microtest)","compute":"memory","nodes":[798,799]},{"id":484,"rule":"w8a16_m32","kernels":["w8a16_gemm_pipelined_m32::w8a16_gemm_pipelined_m32"],"where":"device","numerics":"reference","compute":"tensor_core:mma.sync.m16n8k16.bf16","nodes":[800]},{"id":485,"rule":"gdn_ba_gates_gemv_per_row","kernels":["ssm_preprocess::dense_gemv_ba_gates"],"where":"device","numerics":"reference","compute":"cuda_core","nodes":[801,802]},{"id":486,"rule":"gdn_conv_l2_f32_per_row","kernels":["causal_conv1d::causal_conv1d_update_l2norm_f32"],"where":"device","numerics":"reference","compute":"cuda_core","nodes":[803,804,805]},{"id":487,"rule":"gdn_recurrence_f32_per_row","kernels":["gated_delta_rule::gated_delta_rule_decode_f32"],"where":"device","numerics":"reference","compute":"cuda_core","nodes":[806]},{"id":488,"rule":"gdn_out_norm_f32_per_row","kernels":["norm::gated_rms_norm_f32_input"],"where":"device","numerics":"reference","compute":"memory","nodes":[807]},{"id":489,"rule":"w8a16_m32","kernels":["w8a16_gemm_pipelined_m32::w8a16_gemm_pipelined_m32"],"where":"device","numerics":"reference","compute":"tensor_core:mma.sync.m16n8k16.bf16","nodes":[808]},{"id":490,"rule":"residual_add_post_norm","kernels":["norm::residual_add_rms_norm"],"where":"device","numerics":"reference","compute":"memory","nodes":[809,810]},{"id":491,"rule":"moe_router_gemv","kernels":["gemv::dense_gemv_bf16"],"where":"device","numerics":"reference","compute":"cuda_core","nodes":[811]},{"id":492,"rule":"moe_topk_softmax","kernels":["moe_topk::moe_topk_softmax"],"where":"device","numerics":"reference","compute":"memory","nodes":[812]},{"id":493,"rule":"moe_gate_up_shared_fp8","kernels":["moe_shared_expert_fused_fp8::moe_expert_gate_up_shared_fp8"],"where":"device","numerics":"reference","compute":"cuda_core","nodes":[813,814]},{"id":494,"rule":"moe_silu_down_shared_fp8","kernels":["moe_shared_expert_fused_fp8::moe_expert_silu_down_shared_fp8"],"where":"device","numerics":"reference","compute":"cuda_core","nodes":[815,816,817,818]},{"id":495,"rule":"moe_weighted_sum_blend","kernels":["moe_expert_gemv::moe_weighted_sum_blend"],"where":"device","numerics":"reference","compute":"cuda_core","nodes":[819,820]},{"id":496,"rule":"cross_layer_add_norm","kernels":["residual_add_rms_norm_exact::residual_add_rms_norm_exact"],"where":"device","numerics":"bit_identical(residual_add_rms_norm_exact_microtest)","compute":"memory","nodes":[821,822]},{"id":497,"rule":"w8a16_m32","kernels":["w8a16_gemm_pipelined_m32::w8a16_gemm_pipelined_m32"],"where":"device","numerics":"reference","compute":"tensor_core:mma.sync.m16n8k16.bf16","nodes":[823]},{"id":498,"rule":"deinterleave_qg","kernels":["ssm_preprocess::deinterleave_qg"],"where":"device","numerics":"reference","compute":"memory","nodes":[824]},{"id":499,"rule":"w8a16_m32","kernels":["w8a16_gemm_pipelined_m32::w8a16_gemm_pipelined_m32"],"where":"device","numerics":"reference","compute":"tensor_core:mma.sync.m16n8k16.bf16","nodes":[825]},{"id":500,"rule":"w8a16_m32","kernels":["w8a16_gemm_pipelined_m32::w8a16_gemm_pipelined_m32"],"where":"device","numerics":"reference","compute":"tensor_core:mma.sync.m16n8k16.bf16","nodes":[826]},{"id":501,"rule":"qk_norm_rows","kernels":["norm::rms_norm"],"where":"device","numerics":"reference","compute":"memory","nodes":[827]},{"id":502,"rule":"qk_norm_rows","kernels":["norm::rms_norm"],"where":"device","numerics":"reference","compute":"memory","nodes":[828]},{"id":503,"rule":"rope_mrope_interleaved","kernels":["rope_mrope_interleaved::rope_forward_mrope_interleaved"],"where":"device","numerics":"reference","compute":"memory","nodes":[829]},{"id":504,"rule":"kv_write_bf16","kernels":["reshape_and_cache::reshape_and_cache_flash"],"where":"device","numerics":"reference","compute":"memory","nodes":[830]},{"id":505,"rule":"paged_attention_bf16","kernels":["paged_decode::paged_decode_attn"],"where":"device","numerics":"reference","compute":"cuda_core","nodes":[831]},{"id":506,"rule":"sigmoid_gate_mul","kernels":["residual_add::sigmoid_gate_mul"],"where":"device","numerics":"reference","compute":"memory","nodes":[832]},{"id":507,"rule":"w8a16_m32","kernels":["w8a16_gemm_pipelined_m32::w8a16_gemm_pipelined_m32"],"where":"device","numerics":"reference","compute":"tensor_core:mma.sync.m16n8k16.bf16","nodes":[833]},{"id":508,"rule":"residual_add_post_norm","kernels":["norm::residual_add_rms_norm"],"where":"device","numerics":"reference","compute":"memory","nodes":[834,835]},{"id":509,"rule":"moe_router_gemv","kernels":["gemv::dense_gemv_bf16"],"where":"device","numerics":"reference","compute":"cuda_core","nodes":[836]},{"id":510,"rule":"moe_topk_softmax","kernels":["moe_topk::moe_topk_softmax"],"where":"device","numerics":"reference","compute":"memory","nodes":[837]},{"id":511,"rule":"moe_gate_up_shared_fp8","kernels":["moe_shared_expert_fused_fp8::moe_expert_gate_up_shared_fp8"],"where":"device","numerics":"reference","compute":"cuda_core","nodes":[838,839]},{"id":512,"rule":"moe_silu_down_shared_fp8","kernels":["moe_shared_expert_fused_fp8::moe_expert_silu_down_shared_fp8"],"where":"device","numerics":"reference","compute":"cuda_core","nodes":[840,841,842,843]},{"id":513,"rule":"moe_weighted_sum_blend","kernels":["moe_expert_gemv::moe_weighted_sum_blend"],"where":"device","numerics":"reference","compute":"cuda_core","nodes":[844,845]},{"id":514,"rule":"cross_layer_add_norm","kernels":["residual_add_rms_norm_exact::residual_add_rms_norm_exact"],"where":"device","numerics":"bit_identical(residual_add_rms_norm_exact_microtest)","compute":"memory","nodes":[846,847]},{"id":515,"rule":"w8a16_m32","kernels":["w8a16_gemm_pipelined_m32::w8a16_gemm_pipelined_m32"],"where":"device","numerics":"reference","compute":"tensor_core:mma.sync.m16n8k16.bf16","nodes":[848]},{"id":516,"rule":"gdn_ba_gates_gemv_per_row","kernels":["ssm_preprocess::dense_gemv_ba_gates"],"where":"device","numerics":"reference","compute":"cuda_core","nodes":[849,850]},{"id":517,"rule":"gdn_conv_l2_f32_per_row","kernels":["causal_conv1d::causal_conv1d_update_l2norm_f32"],"where":"device","numerics":"reference","compute":"cuda_core","nodes":[851,852,853]},{"id":518,"rule":"gdn_recurrence_f32_per_row","kernels":["gated_delta_rule::gated_delta_rule_decode_f32"],"where":"device","numerics":"reference","compute":"cuda_core","nodes":[854]},{"id":519,"rule":"gdn_out_norm_f32_per_row","kernels":["norm::gated_rms_norm_f32_input"],"where":"device","numerics":"reference","compute":"memory","nodes":[855]},{"id":520,"rule":"w8a16_m32","kernels":["w8a16_gemm_pipelined_m32::w8a16_gemm_pipelined_m32"],"where":"device","numerics":"reference","compute":"tensor_core:mma.sync.m16n8k16.bf16","nodes":[856]},{"id":521,"rule":"residual_add_post_norm","kernels":["norm::residual_add_rms_norm"],"where":"device","numerics":"reference","compute":"memory","nodes":[857,858]},{"id":522,"rule":"moe_router_gemv","kernels":["gemv::dense_gemv_bf16"],"where":"device","numerics":"reference","compute":"cuda_core","nodes":[859]},{"id":523,"rule":"moe_topk_softmax","kernels":["moe_topk::moe_topk_softmax"],"where":"device","numerics":"reference","compute":"memory","nodes":[860]},{"id":524,"rule":"moe_gate_up_shared_fp8","kernels":["moe_shared_expert_fused_fp8::moe_expert_gate_up_shared_fp8"],"where":"device","numerics":"reference","compute":"cuda_core","nodes":[861,862]},{"id":525,"rule":"moe_silu_down_shared_fp8","kernels":["moe_shared_expert_fused_fp8::moe_expert_silu_down_shared_fp8"],"where":"device","numerics":"reference","compute":"cuda_core","nodes":[863,864,865,866]},{"id":526,"rule":"moe_weighted_sum_blend","kernels":["moe_expert_gemv::moe_weighted_sum_blend"],"where":"device","numerics":"reference","compute":"cuda_core","nodes":[867,868]},{"id":527,"rule":"cross_layer_add_norm","kernels":["residual_add_rms_norm_exact::residual_add_rms_norm_exact"],"where":"device","numerics":"bit_identical(residual_add_rms_norm_exact_microtest)","compute":"memory","nodes":[869,870]},{"id":528,"rule":"w8a16_m32","kernels":["w8a16_gemm_pipelined_m32::w8a16_gemm_pipelined_m32"],"where":"device","numerics":"reference","compute":"tensor_core:mma.sync.m16n8k16.bf16","nodes":[871]},{"id":529,"rule":"gdn_ba_gates_gemv_per_row","kernels":["ssm_preprocess::dense_gemv_ba_gates"],"where":"device","numerics":"reference","compute":"cuda_core","nodes":[872,873]},{"id":530,"rule":"gdn_conv_l2_f32_per_row","kernels":["causal_conv1d::causal_conv1d_update_l2norm_f32"],"where":"device","numerics":"reference","compute":"cuda_core","nodes":[874,875,876]},{"id":531,"rule":"gdn_recurrence_f32_per_row","kernels":["gated_delta_rule::gated_delta_rule_decode_f32"],"where":"device","numerics":"reference","compute":"cuda_core","nodes":[877]},{"id":532,"rule":"gdn_out_norm_f32_per_row","kernels":["norm::gated_rms_norm_f32_input"],"where":"device","numerics":"reference","compute":"memory","nodes":[878]},{"id":533,"rule":"w8a16_m32","kernels":["w8a16_gemm_pipelined_m32::w8a16_gemm_pipelined_m32"],"where":"device","numerics":"reference","compute":"tensor_core:mma.sync.m16n8k16.bf16","nodes":[879]},{"id":534,"rule":"residual_add_post_norm","kernels":["norm::residual_add_rms_norm"],"where":"device","numerics":"reference","compute":"memory","nodes":[880,881]},{"id":535,"rule":"moe_router_gemv","kernels":["gemv::dense_gemv_bf16"],"where":"device","numerics":"reference","compute":"cuda_core","nodes":[882]},{"id":536,"rule":"moe_topk_softmax","kernels":["moe_topk::moe_topk_softmax"],"where":"device","numerics":"reference","compute":"memory","nodes":[883]},{"id":537,"rule":"moe_gate_up_shared_fp8","kernels":["moe_shared_expert_fused_fp8::moe_expert_gate_up_shared_fp8"],"where":"device","numerics":"reference","compute":"cuda_core","nodes":[884,885]},{"id":538,"rule":"moe_silu_down_shared_fp8","kernels":["moe_shared_expert_fused_fp8::moe_expert_silu_down_shared_fp8"],"where":"device","numerics":"reference","compute":"cuda_core","nodes":[886,887,888,889]},{"id":539,"rule":"moe_weighted_sum_blend","kernels":["moe_expert_gemv::moe_weighted_sum_blend"],"where":"device","numerics":"reference","compute":"cuda_core","nodes":[890,891]},{"id":540,"rule":"cross_layer_add_norm","kernels":["residual_add_rms_norm_exact::residual_add_rms_norm_exact"],"where":"device","numerics":"bit_identical(residual_add_rms_norm_exact_microtest)","compute":"memory","nodes":[892,893]},{"id":541,"rule":"w8a16_m32","kernels":["w8a16_gemm_pipelined_m32::w8a16_gemm_pipelined_m32"],"where":"device","numerics":"reference","compute":"tensor_core:mma.sync.m16n8k16.bf16","nodes":[894]},{"id":542,"rule":"gdn_ba_gates_gemv_per_row","kernels":["ssm_preprocess::dense_gemv_ba_gates"],"where":"device","numerics":"reference","compute":"cuda_core","nodes":[895,896]},{"id":543,"rule":"gdn_conv_l2_f32_per_row","kernels":["causal_conv1d::causal_conv1d_update_l2norm_f32"],"where":"device","numerics":"reference","compute":"cuda_core","nodes":[897,898,899]},{"id":544,"rule":"gdn_recurrence_f32_per_row","kernels":["gated_delta_rule::gated_delta_rule_decode_f32"],"where":"device","numerics":"reference","compute":"cuda_core","nodes":[900]},{"id":545,"rule":"gdn_out_norm_f32_per_row","kernels":["norm::gated_rms_norm_f32_input"],"where":"device","numerics":"reference","compute":"memory","nodes":[901]},{"id":546,"rule":"w8a16_m32","kernels":["w8a16_gemm_pipelined_m32::w8a16_gemm_pipelined_m32"],"where":"device","numerics":"reference","compute":"tensor_core:mma.sync.m16n8k16.bf16","nodes":[902]},{"id":547,"rule":"residual_add_post_norm","kernels":["norm::residual_add_rms_norm"],"where":"device","numerics":"reference","compute":"memory","nodes":[903,904]},{"id":548,"rule":"moe_router_gemv","kernels":["gemv::dense_gemv_bf16"],"where":"device","numerics":"reference","compute":"cuda_core","nodes":[905]},{"id":549,"rule":"moe_topk_softmax","kernels":["moe_topk::moe_topk_softmax"],"where":"device","numerics":"reference","compute":"memory","nodes":[906]},{"id":550,"rule":"moe_gate_up_shared_fp8","kernels":["moe_shared_expert_fused_fp8::moe_expert_gate_up_shared_fp8"],"where":"device","numerics":"reference","compute":"cuda_core","nodes":[907,908]},{"id":551,"rule":"moe_silu_down_shared_fp8","kernels":["moe_shared_expert_fused_fp8::moe_expert_silu_down_shared_fp8"],"where":"device","numerics":"reference","compute":"cuda_core","nodes":[909,910,911,912]},{"id":552,"rule":"moe_weighted_sum_blend","kernels":["moe_expert_gemv::moe_weighted_sum_blend"],"where":"device","numerics":"reference","compute":"cuda_core","nodes":[913,914]},{"id":553,"rule":"cross_layer_add_norm","kernels":["residual_add_rms_norm_exact::residual_add_rms_norm_exact"],"where":"device","numerics":"bit_identical(residual_add_rms_norm_exact_microtest)","compute":"memory","nodes":[915,916]},{"id":554,"rule":"w8a16_m32","kernels":["w8a16_gemm_pipelined_m32::w8a16_gemm_pipelined_m32"],"where":"device","numerics":"reference","compute":"tensor_core:mma.sync.m16n8k16.bf16","nodes":[917]},{"id":555,"rule":"deinterleave_qg","kernels":["ssm_preprocess::deinterleave_qg"],"where":"device","numerics":"reference","compute":"memory","nodes":[918]},{"id":556,"rule":"w8a16_m32","kernels":["w8a16_gemm_pipelined_m32::w8a16_gemm_pipelined_m32"],"where":"device","numerics":"reference","compute":"tensor_core:mma.sync.m16n8k16.bf16","nodes":[919]},{"id":557,"rule":"w8a16_m32","kernels":["w8a16_gemm_pipelined_m32::w8a16_gemm_pipelined_m32"],"where":"device","numerics":"reference","compute":"tensor_core:mma.sync.m16n8k16.bf16","nodes":[920]},{"id":558,"rule":"qk_norm_rows","kernels":["norm::rms_norm"],"where":"device","numerics":"reference","compute":"memory","nodes":[921]},{"id":559,"rule":"qk_norm_rows","kernels":["norm::rms_norm"],"where":"device","numerics":"reference","compute":"memory","nodes":[922]},{"id":560,"rule":"rope_mrope_interleaved","kernels":["rope_mrope_interleaved::rope_forward_mrope_interleaved"],"where":"device","numerics":"reference","compute":"memory","nodes":[923]},{"id":561,"rule":"kv_write_bf16","kernels":["reshape_and_cache::reshape_and_cache_flash"],"where":"device","numerics":"reference","compute":"memory","nodes":[924]},{"id":562,"rule":"paged_attention_bf16","kernels":["paged_decode::paged_decode_attn"],"where":"device","numerics":"reference","compute":"cuda_core","nodes":[925]},{"id":563,"rule":"sigmoid_gate_mul","kernels":["residual_add::sigmoid_gate_mul"],"where":"device","numerics":"reference","compute":"memory","nodes":[926]},{"id":564,"rule":"w8a16_m32","kernels":["w8a16_gemm_pipelined_m32::w8a16_gemm_pipelined_m32"],"where":"device","numerics":"reference","compute":"tensor_core:mma.sync.m16n8k16.bf16","nodes":[927]},{"id":565,"rule":"residual_add_post_norm","kernels":["norm::residual_add_rms_norm"],"where":"device","numerics":"reference","compute":"memory","nodes":[928,929]},{"id":566,"rule":"moe_router_gemv","kernels":["gemv::dense_gemv_bf16"],"where":"device","numerics":"reference","compute":"cuda_core","nodes":[930]},{"id":567,"rule":"moe_topk_softmax","kernels":["moe_topk::moe_topk_softmax"],"where":"device","numerics":"reference","compute":"memory","nodes":[931]},{"id":568,"rule":"moe_gate_up_shared_fp8","kernels":["moe_shared_expert_fused_fp8::moe_expert_gate_up_shared_fp8"],"where":"device","numerics":"reference","compute":"cuda_core","nodes":[932,933]},{"id":569,"rule":"moe_silu_down_shared_fp8","kernels":["moe_shared_expert_fused_fp8::moe_expert_silu_down_shared_fp8"],"where":"device","numerics":"reference","compute":"cuda_core","nodes":[934,935,936,937]},{"id":570,"rule":"moe_weighted_sum_blend","kernels":["moe_expert_gemv::moe_weighted_sum_blend"],"where":"device","numerics":"reference","compute":"cuda_core","nodes":[938,939]},{"id":571,"rule":"ffn_residual_add","kernels":["residual_add::bf16_residual_add"],"where":"device","numerics":"reference","compute":"memory","nodes":[940]},{"id":572,"rule":"final_norm","kernels":["norm::rms_norm"],"where":"device","numerics":"reference","compute":"memory","nodes":[941]},{"id":573,"rule":"lm_head_bf16_gemv","kernels":["gemv::dense_gemv_bf16"],"where":"device","numerics":"reference","compute":"cuda_core","nodes":[942]},{"id":574,"rule":"argmax_host","kernels":[],"where":"host","numerics":"reference","compute":null,"nodes":[943]}],"collapsed":{"prologue":["embed"],"layout":["gdn+moe_ffn","gdn+moe_ffn","gdn+moe_ffn","attn+moe_ffn","gdn+moe_ffn","gdn+moe_ffn","gdn+moe_ffn","attn+moe_ffn","gdn+moe_ffn","gdn+moe_ffn","gdn+moe_ffn","attn+moe_ffn","gdn+moe_ffn","gdn+moe_ffn","gdn+moe_ffn","attn+moe_ffn","gdn+moe_ffn","gdn+moe_ffn","gdn+moe_ffn","attn+moe_ffn","gdn+moe_ffn","gdn+moe_ffn","gdn+moe_ffn","attn+moe_ffn","gdn+moe_ffn","gdn+moe_ffn","gdn+moe_ffn","attn+moe_ffn","gdn+moe_ffn","gdn+moe_ffn","gdn+moe_ffn","attn+moe_ffn","gdn+moe_ffn","gdn+moe_ffn","gdn+moe_ffn","attn+moe_ffn","gdn+moe_ffn","gdn+moe_ffn","gdn+moe_ffn","attn+moe_ffn"],"epilogue":["head"],"templates":{"embed":{"blocks":["embed"],"nodes":[{"id":"embed.embed","op":"embed","kind":null,"block_type":"embed","precision":{"weight":null,"activation":null,"output":"bf16"},"kernel":null,"family":null,"compute":null,"where":"host","group":0,"inserted":false}],"edges":[],"ports":{"in":[],"out":[0]},"instances":[]},"gdn+moe_ffn":{"blocks":["gdn","moe_ffn"],"nodes":[{"id":"gdn.input_norm","op":"rms_norm","kind":"linear_attention","block_type":"gdn","precision":{"weight":null,"activation":"bf16","output":"bf16"},"kernel":"norm::rms_norm_residual","family":"rms_norm","compute":"memory","where":"device","group":0,"inserted":false},{"id":"gdn.qkvz","op":"linear:qkvz","kind":"linear_attention","block_type":"gdn","precision":{"weight":"fp8/block128x128","activation":"bf16","output":"bf16"},"kernel":"w8a16_gemm_pipelined_m32::w8a16_gemm_pipelined_m32","family":"w8a16_gemm","compute":"tensor_core:mma.sync.m16n8k16.bf16","where":"device","group":1,"inserted":false},{"id":"gdn.ba","op":"linear:ba","kind":"linear_attention","block_type":"gdn","precision":{"weight":"bf16","activation":"bf16","output":"f32"},"kernel":"ssm_preprocess::dense_gemv_ba_gates","family":"gdn_ba_gates_gemv","compute":"cuda_core","where":"device","group":2,"inserted":false},{"id":"gdn.gates","op":"gdn_gates","kind":"linear_attention","block_type":"gdn","precision":{"weight":null,"activation":"f32","output":"f32,f32"},"kernel":"ssm_preprocess::dense_gemv_ba_gates","family":"gdn_ba_gates_gemv","compute":"cuda_core","where":"device","group":2,"inserted":false},{"id":"gdn.conv","op":"conv1d_update","kind":"linear_attention","block_type":"gdn","precision":{"weight":null,"activation":"bf16","output":"f32"},"kernel":"causal_conv1d::causal_conv1d_update_l2norm_f32","family":"causal_conv1d_l2norm","compute":"cuda_core","where":"device","group":3,"inserted":false},{"id":"gdn.conv_ckpt","op":"state_snapshot","kind":"linear_attention","block_type":"gdn","precision":{"weight":null,"activation":"f32","output":null},"kernel":"causal_conv1d::causal_conv1d_update_l2norm_f32","family":"causal_conv1d_l2norm","compute":"cuda_core","where":"device","group":3,"inserted":false},{"id":"gdn.l2","op":"l2_norm","kind":"linear_attention","block_type":"gdn","precision":{"weight":null,"activation":"f32","output":"f32"},"kernel":"causal_conv1d::causal_conv1d_update_l2norm_f32","family":"causal_conv1d_l2norm","compute":"cuda_core","where":"device","group":3,"inserted":false},{"id":"gdn.recur","op":"gdn_recurrence","kind":"linear_attention","block_type":"gdn","precision":{"weight":null,"activation":"f32","output":"f32"},"kernel":"gated_delta_rule::gated_delta_rule_decode_f32","family":"gdn_recurrence","compute":"cuda_core","where":"device","group":4,"inserted":false},{"id":"gdn.out_norm","op":"gated_rms_norm","kind":"linear_attention","block_type":"gdn","precision":{"weight":null,"activation":"f32","output":"bf16"},"kernel":"norm::gated_rms_norm_f32_input","family":"gated_rms_norm","compute":"memory","where":"device","group":5,"inserted":false},{"id":"gdn.out","op":"linear:gdn_out","kind":"linear_attention","block_type":"gdn","precision":{"weight":"fp8/block128x128","activation":"bf16","output":"bf16"},"kernel":"w8a16_gemm_pipelined_m32::w8a16_gemm_pipelined_m32","family":"w8a16_gemm","compute":"tensor_core:mma.sync.m16n8k16.bf16","where":"device","group":6,"inserted":false},{"id":"gdn.add","op":"residual_add","kind":"linear_attention","block_type":"gdn","precision":{"weight":null,"activation":"bf16","output":"bf16"},"kernel":"norm::residual_add_rms_norm","family":"rms_norm","compute":"memory","where":"device","group":7,"inserted":false},{"id":"moe_ffn.post_norm","op":"rms_norm","kind":"linear_attention","block_type":"moe_ffn","precision":{"weight":null,"activation":"f32","output":"bf16"},"kernel":"norm::residual_add_rms_norm","family":"rms_norm","compute":"memory","where":"device","group":7,"inserted":false},{"id":"moe_ffn.router","op":"router","kind":"linear_attention","block_type":"moe_ffn","precision":{"weight":"bf16","activation":"bf16","output":"bf16"},"kernel":"gemv::dense_gemv_bf16","family":"dense_bf16","compute":"cuda_core","where":"device","group":8,"inserted":false},{"id":"moe_ffn.top_k","op":"top_k","kind":"linear_attention","block_type":"moe_ffn","precision":{"weight":null,"activation":"bf16","output":"f32,i32"},"kernel":"moe_topk::moe_topk_softmax","family":"moe_topk","compute":"memory","where":"device","group":9,"inserted":false},{"id":"moe_ffn.experts_gate_up","op":"expert_gate_up","kind":"linear_attention","block_type":"moe_ffn","precision":{"weight":"fp8/block128x128","activation":"bf16","output":"bf16"},"kernel":"moe_shared_expert_fused_fp8::moe_expert_gate_up_shared_fp8","family":"moe_fp8_1row","compute":"cuda_core","where":"device","group":10,"inserted":false},{"id":"moe_ffn.shared_gate_up","op":"linear:shared_gate_up","kind":"linear_attention","block_type":"moe_ffn","precision":{"weight":"fp8/block128x128","activation":"bf16","output":"bf16"},"kernel":"moe_shared_expert_fused_fp8::moe_expert_gate_up_shared_fp8","family":"moe_fp8_1row","compute":"cuda_core","where":"device","group":10,"inserted":false},{"id":"moe_ffn.experts_act","op":"silu_mul","kind":"linear_attention","block_type":"moe_ffn","precision":{"weight":null,"activation":"bf16","output":"f32"},"kernel":"moe_shared_expert_fused_fp8::moe_expert_silu_down_shared_fp8","family":"moe_fp8_1row","compute":"cuda_core","where":"device","group":11,"inserted":false},{"id":"moe_ffn.experts_down","op":"expert_down","kind":"linear_attention","block_type":"moe_ffn","precision":{"weight":"fp8/block128x128","activation":"f32","output":"bf16"},"kernel":"moe_shared_expert_fused_fp8::moe_expert_silu_down_shared_fp8","family":"moe_fp8_1row","compute":"cuda_core","where":"device","group":11,"inserted":false},{"id":"moe_ffn.shared_act","op":"silu_mul","kind":"linear_attention","block_type":"moe_ffn","precision":{"weight":null,"activation":"bf16","output":"f32"},"kernel":"moe_shared_expert_fused_fp8::moe_expert_silu_down_shared_fp8","family":"moe_fp8_1row","compute":"cuda_core","where":"device","group":11,"inserted":false},{"id":"moe_ffn.shared_down","op":"linear:shared_down","kind":"linear_attention","block_type":"moe_ffn","precision":{"weight":"fp8/block128x128","activation":"f32","output":"bf16"},"kernel":"moe_shared_expert_fused_fp8::moe_expert_silu_down_shared_fp8","family":"moe_fp8_1row","compute":"cuda_core","where":"device","group":11,"inserted":false},{"id":"moe_ffn.shared_gate","op":"linear:shared_gate","kind":"linear_attention","block_type":"moe_ffn","precision":{"weight":"bf16","activation":"bf16","output":"f32"},"kernel":"moe_expert_gemv::moe_weighted_sum_blend","family":"moe_blend","compute":"cuda_core","where":"device","group":12,"inserted":false},{"id":"moe_ffn.blend","op":"blend","kind":"linear_attention","block_type":"moe_ffn","precision":{"weight":null,"activation":"bf16","output":"bf16"},"kernel":"moe_expert_gemv::moe_weighted_sum_blend","family":"moe_blend","compute":"cuda_core","where":"device","group":12,"inserted":false},{"id":"moe_ffn.add","op":"residual_add","kind":"linear_attention","block_type":"moe_ffn","precision":{"weight":null,"activation":"bf16","output":"bf16"},"kernel":"residual_add_rms_norm_exact::residual_add_rms_norm_exact","family":"rms_norm","compute":"memory","where":"device","group":13,"inserted":false}],"edges":[{"from":0,"to":1,"format":"bf16","fused":false},{"from":0,"to":2,"format":"bf16","fused":false},{"from":1,"to":4,"format":"bf16","fused":false},{"from":1,"to":8,"format":"bf16","fused":false},{"from":2,"to":3,"format":"f32","fused":true},{"from":3,"to":7,"format":"f32","fused":false},{"from":3,"to":7,"format":"f32","fused":false},{"from":4,"to":5,"format":"f32","fused":true},{"from":4,"to":6,"format":"f32","fused":true},{"from":6,"to":7,"format":"f32","fused":false},{"from":7,"to":8,"format":"f32","fused":false},{"from":8,"to":9,"format":"bf16","fused":false},{"from":9,"to":10,"format":"bf16","fused":false},{"from":10,"to":11,"format":"bf16","fused":false},{"from":10,"to":22,"format":"bf16","fused":false},{"from":11,"to":12,"format":"bf16","fused":false},{"from":11,"to":14,"format":"bf16","fused":false},{"from":11,"to":15,"format":"bf16","fused":false},{"from":11,"to":20,"format":"bf16","fused":false},{"from":12,"to":13,"format":"bf16","fused":false},{"from":13,"to":21,"format":"f32","fused":false},{"from":13,"to":14,"format":"i32","fused":false},{"from":13,"to":17,"format":"i32","fused":false},{"from":14,"to":16,"format":"bf16","fused":false},{"from":15,"to":18,"format":"bf16","fused":false},{"from":16,"to":17,"format":"f32","fused":true},{"from":17,"to":21,"format":"bf16","fused":false},{"from":18,"to":19,"format":"f32","fused":true},{"from":19,"to":21,"format":"bf16","fused":false},{"from":20,"to":21,"format":"f32","fused":true},{"from":21,"to":22,"format":"bf16","fused":false}],"ports":{"in":[0,10],"out":[22]},"instances":[0,1,2,4,5,6,8,9,10,12,13,14,16,17,18,20,21,22,24,25,26,28,29,30,32,33,34,36,37,38]},"attn+moe_ffn":{"blocks":["attn","moe_ffn"],"nodes":[{"id":"attn.input_norm","op":"rms_norm","kind":"full_attention","block_type":"attn","precision":{"weight":null,"activation":"bf16","output":"bf16"},"kernel":"residual_add_rms_norm_exact::residual_add_rms_norm_exact","family":"rms_norm","compute":"memory","where":"device","group":0,"inserted":false},{"id":"attn.q","op":"linear:q","kind":"full_attention","block_type":"attn","precision":{"weight":"fp8/block128x128","activation":"bf16","output":"bf16"},"kernel":"w8a16_gemm_pipelined_m32::w8a16_gemm_pipelined_m32","family":"w8a16_gemm","compute":"tensor_core:mma.sync.m16n8k16.bf16","where":"device","group":1,"inserted":false},{"id":"attn.q_split","op":"split","kind":"full_attention","block_type":"attn","precision":{"weight":null,"activation":"bf16","output":"bf16,bf16"},"kernel":"ssm_preprocess::deinterleave_qg","family":"deinterleave_qg","compute":"memory","where":"device","group":2,"inserted":false},{"id":"attn.k","op":"linear:k","kind":"full_attention","block_type":"attn","precision":{"weight":"fp8/block128x128","activation":"bf16","output":"bf16"},"kernel":"w8a16_gemm_pipelined_m32::w8a16_gemm_pipelined_m32","family":"w8a16_gemm","compute":"tensor_core:mma.sync.m16n8k16.bf16","where":"device","group":3,"inserted":false},{"id":"attn.v","op":"linear:v","kind":"full_attention","block_type":"attn","precision":{"weight":"fp8/block128x128","activation":"bf16","output":"bf16"},"kernel":"w8a16_gemm_pipelined_m32::w8a16_gemm_pipelined_m32","family":"w8a16_gemm","compute":"tensor_core:mma.sync.m16n8k16.bf16","where":"device","group":4,"inserted":false},{"id":"attn.q_norm","op":"qk_norm","kind":"full_attention","block_type":"attn","precision":{"weight":null,"activation":"bf16","output":"bf16"},"kernel":"norm::rms_norm","family":"rms_norm","compute":"memory","where":"device","group":5,"inserted":false},{"id":"attn.k_norm","op":"qk_norm","kind":"full_attention","block_type":"attn","precision":{"weight":null,"activation":"bf16","output":"bf16"},"kernel":"norm::rms_norm","family":"rms_norm","compute":"memory","where":"device","group":6,"inserted":false},{"id":"attn.rope","op":"rope","kind":"full_attention","block_type":"attn","precision":{"weight":null,"activation":"bf16","output":"bf16,bf16"},"kernel":"rope_mrope_interleaved::rope_forward_mrope_interleaved","family":"rope","compute":"memory","where":"device","group":7,"inserted":false},{"id":"attn.kv_write","op":"kv_write","kind":"full_attention","block_type":"attn","precision":{"weight":null,"activation":"bf16","output":null},"kernel":"reshape_and_cache::reshape_and_cache_flash","family":"kv_write","compute":"memory","where":"device","group":8,"inserted":false},{"id":"attn.attend","op":"paged_attention","kind":"full_attention","block_type":"attn","precision":{"weight":null,"activation":"bf16","output":"bf16"},"kernel":"paged_decode::paged_decode_attn","family":"paged_decode_attn","compute":"cuda_core","where":"device","group":9,"inserted":false},{"id":"attn.gate_mul","op":"sigmoid_gate_mul","kind":"full_attention","block_type":"attn","precision":{"weight":null,"activation":"bf16","output":"bf16"},"kernel":"residual_add::sigmoid_gate_mul","family":"sigmoid_gate_mul","compute":"memory","where":"device","group":10,"inserted":false},{"id":"attn.o","op":"linear:o","kind":"full_attention","block_type":"attn","precision":{"weight":"fp8/block128x128","activation":"bf16","output":"bf16"},"kernel":"w8a16_gemm_pipelined_m32::w8a16_gemm_pipelined_m32","family":"w8a16_gemm","compute":"tensor_core:mma.sync.m16n8k16.bf16","where":"device","group":11,"inserted":false},{"id":"attn.add","op":"residual_add","kind":"full_attention","block_type":"attn","precision":{"weight":null,"activation":"bf16","output":"bf16"},"kernel":"norm::residual_add_rms_norm","family":"rms_norm","compute":"memory","where":"device","group":12,"inserted":false},{"id":"moe_ffn.post_norm","op":"rms_norm","kind":"full_attention","block_type":"moe_ffn","precision":{"weight":null,"activation":"f32","output":"bf16"},"kernel":"norm::residual_add_rms_norm","family":"rms_norm","compute":"memory","where":"device","group":12,"inserted":false},{"id":"moe_ffn.router","op":"router","kind":"full_attention","block_type":"moe_ffn","precision":{"weight":"bf16","activation":"bf16","output":"bf16"},"kernel":"gemv::dense_gemv_bf16","family":"dense_bf16","compute":"cuda_core","where":"device","group":13,"inserted":false},{"id":"moe_ffn.top_k","op":"top_k","kind":"full_attention","block_type":"moe_ffn","precision":{"weight":null,"activation":"bf16","output":"f32,i32"},"kernel":"moe_topk::moe_topk_softmax","family":"moe_topk","compute":"memory","where":"device","group":14,"inserted":false},{"id":"moe_ffn.experts_gate_up","op":"expert_gate_up","kind":"full_attention","block_type":"moe_ffn","precision":{"weight":"fp8/block128x128","activation":"bf16","output":"bf16"},"kernel":"moe_shared_expert_fused_fp8::moe_expert_gate_up_shared_fp8","family":"moe_fp8_1row","compute":"cuda_core","where":"device","group":15,"inserted":false},{"id":"moe_ffn.shared_gate_up","op":"linear:shared_gate_up","kind":"full_attention","block_type":"moe_ffn","precision":{"weight":"fp8/block128x128","activation":"bf16","output":"bf16"},"kernel":"moe_shared_expert_fused_fp8::moe_expert_gate_up_shared_fp8","family":"moe_fp8_1row","compute":"cuda_core","where":"device","group":15,"inserted":false},{"id":"moe_ffn.experts_act","op":"silu_mul","kind":"full_attention","block_type":"moe_ffn","precision":{"weight":null,"activation":"bf16","output":"f32"},"kernel":"moe_shared_expert_fused_fp8::moe_expert_silu_down_shared_fp8","family":"moe_fp8_1row","compute":"cuda_core","where":"device","group":16,"inserted":false},{"id":"moe_ffn.experts_down","op":"expert_down","kind":"full_attention","block_type":"moe_ffn","precision":{"weight":"fp8/block128x128","activation":"f32","output":"bf16"},"kernel":"moe_shared_expert_fused_fp8::moe_expert_silu_down_shared_fp8","family":"moe_fp8_1row","compute":"cuda_core","where":"device","group":16,"inserted":false},{"id":"moe_ffn.shared_act","op":"silu_mul","kind":"full_attention","block_type":"moe_ffn","precision":{"weight":null,"activation":"bf16","output":"f32"},"kernel":"moe_shared_expert_fused_fp8::moe_expert_silu_down_shared_fp8","family":"moe_fp8_1row","compute":"cuda_core","where":"device","group":16,"inserted":false},{"id":"moe_ffn.shared_down","op":"linear:shared_down","kind":"full_attention","block_type":"moe_ffn","precision":{"weight":"fp8/block128x128","activation":"f32","output":"bf16"},"kernel":"moe_shared_expert_fused_fp8::moe_expert_silu_down_shared_fp8","family":"moe_fp8_1row","compute":"cuda_core","where":"device","group":16,"inserted":false},{"id":"moe_ffn.shared_gate","op":"linear:shared_gate","kind":"full_attention","block_type":"moe_ffn","precision":{"weight":"bf16","activation":"bf16","output":"f32"},"kernel":"moe_expert_gemv::moe_weighted_sum_blend","family":"moe_blend","compute":"cuda_core","where":"device","group":17,"inserted":false},{"id":"moe_ffn.blend","op":"blend","kind":"full_attention","block_type":"moe_ffn","precision":{"weight":null,"activation":"bf16","output":"bf16"},"kernel":"moe_expert_gemv::moe_weighted_sum_blend","family":"moe_blend","compute":"cuda_core","where":"device","group":17,"inserted":false},{"id":"moe_ffn.add","op":"residual_add","kind":"full_attention","block_type":"moe_ffn","precision":{"weight":null,"activation":"bf16","output":"bf16"},"kernel":"residual_add_rms_norm_exact::residual_add_rms_norm_exact","family":"rms_norm","compute":"memory","where":"device","group":18,"inserted":false}],"edges":[{"from":0,"to":1,"format":"bf16","fused":false},{"from":0,"to":3,"format":"bf16","fused":false},{"from":0,"to":4,"format":"bf16","fused":false},{"from":1,"to":2,"format":"bf16","fused":false},{"from":2,"to":5,"format":"bf16","fused":false},{"from":2,"to":10,"format":"bf16","fused":false},{"from":3,"to":6,"format":"bf16","fused":false},{"from":4,"to":8,"format":"bf16","fused":false},{"from":4,"to":9,"format":"bf16","fused":false},{"from":5,"to":7,"format":"bf16","fused":false},{"from":6,"to":7,"format":"bf16","fused":false},{"from":7,"to":9,"format":"bf16","fused":false},{"from":7,"to":8,"format":"bf16","fused":false},{"from":7,"to":9,"format":"bf16","fused":false},{"from":9,"to":10,"format":"bf16","fused":false},{"from":10,"to":11,"format":"bf16","fused":false},{"from":11,"to":12,"format":"bf16","fused":false},{"from":12,"to":13,"format":"bf16","fused":false},{"from":12,"to":24,"format":"bf16","fused":false},{"from":13,"to":14,"format":"bf16","fused":false},{"from":13,"to":16,"format":"bf16","fused":false},{"from":13,"to":17,"format":"bf16","fused":false},{"from":13,"to":22,"format":"bf16","fused":false},{"from":14,"to":15,"format":"bf16","fused":false},{"from":15,"to":23,"format":"f32","fused":false},{"from":15,"to":16,"format":"i32","fused":false},{"from":15,"to":19,"format":"i32","fused":false},{"from":16,"to":18,"format":"bf16","fused":false},{"from":17,"to":20,"format":"bf16","fused":false},{"from":18,"to":19,"format":"f32","fused":true},{"from":19,"to":23,"format":"bf16","fused":false},{"from":20,"to":21,"format":"f32","fused":true},{"from":21,"to":23,"format":"bf16","fused":false},{"from":22,"to":23,"format":"f32","fused":true},{"from":23,"to":24,"format":"bf16","fused":false}],"ports":{"in":[0,12],"out":[24]},"instances":[3,7,11,15,19,23,27,31,35,39]},"head":{"blocks":["head"],"nodes":[{"id":"head.final_norm","op":"final_norm","kind":null,"block_type":"head","precision":{"weight":null,"activation":"bf16","output":"bf16"},"kernel":"norm::rms_norm","family":"rms_norm","compute":"memory","where":"device","group":0,"inserted":false},{"id":"head.lm_head","op":"lm_head","kind":null,"block_type":"head","precision":{"weight":"bf16","activation":"bf16","output":"bf16"},"kernel":"gemv::dense_gemv_bf16","family":"dense_bf16","compute":"cuda_core","where":"device","group":1,"inserted":false},{"id":"head.argmax","op":"argmax","kind":null,"block_type":"head","precision":{"weight":null,"activation":"bf16","output":"i32"},"kernel":null,"family":null,"compute":null,"where":"host","group":2,"inserted":false}],"edges":[{"from":0,"to":1,"format":"bf16","fused":false},{"from":1,"to":2,"format":"bf16","fused":false}],"ports":{"in":[0],"out":[]},"instances":[]}}},"provenance":{"kind":"preview","engine_commit":"c5b9a80e635df7c1056a5b617272970f25daba64","planner_sha256":"f4735654ee1d5a9bf4f56e836969140d5bfb26231104bf13118103c69ee8b156","planner_built_from":"c5b9a80e635df7c1056a5b617272970f25daba64","planner_header":{"recipe":"qwen3.6/qwen3.6-35b-a3b-fp8-bf16head","checkpoint":"Qwen/Qwen3.6-35B-A3B-FP8","target":"gb10/qwen3.6-35b-a3b/nvfp4","settings":"activation_quantization=adaptive decode_split_silu=on gemv_sw=on kv_cache_dtype=bf16 lm_head_dtype=bf16 rms_norm_act_quant=off row_tiers=canonical ssm_ba_gates_hopper=on ssm_batched_recurrent=on ssm_h_dtype=f32 w4a16_tc=on","opt-in levers":"none"},"commands":["met circuit plan --root <tree> --checkpoint qwen3.6/qwen3.6-35b-a3b-fp8-bf16head --hardware gb10 --precision recipe --format plan --mode decode --rows 1","met circuit plan --root <tree> --checkpoint qwen3.6/qwen3.6-35b-a3b-fp8-bf16head --hardware gb10 --precision recipe --format plan --mode multi_seq --rows 2","met circuit plan --root <tree> --checkpoint qwen3.6/qwen3.6-35b-a3b-fp8-bf16head --hardware gb10 --precision recipe --format plan --mode multi_seq --rows 4","met circuit plan --root <tree> --checkpoint qwen3.6/qwen3.6-35b-a3b-fp8-bf16head --hardware gb10 --precision recipe --format plan --mode multi_seq --rows 8","met circuit plan --root <tree> --checkpoint qwen3.6/qwen3.6-35b-a3b-fp8-bf16head --hardware gb10 --precision recipe --format plan --mode multi_seq --rows 12","met circuit plan --root <tree> --checkpoint qwen3.6/qwen3.6-35b-a3b-fp8-bf16head --hardware gb10 --precision recipe --format plan --mode multi_seq --rows 16","met circuit plan --root <tree> --checkpoint qwen3.6/qwen3.6-35b-a3b-fp8-bf16head --hardware gb10 --precision recipe --format plan --mode multi_seq --rows 24","met circuit plan --root <tree> --checkpoint qwen3.6/qwen3.6-35b-a3b-fp8-bf16head --hardware gb10 --precision recipe --format plan --mode multi_seq --rows 32","met circuit plan --root <tree> --checkpoint qwen3.6/qwen3.6-35b-a3b-fp8-bf16head --hardware gb10 --precision recipe --format plan --mode multi_seq --rows 48","met circuit plan --root <tree> --checkpoint qwen3.6/qwen3.6-35b-a3b-fp8-bf16head --hardware gb10 --precision recipe --format plan --mode multi_seq --rows 64","met circuit plan --root <tree> --checkpoint qwen3.6/qwen3.6-35b-a3b-fp8-bf16head --hardware gb10 --precision recipe --format plan --mode multi_seq --rows 96","met circuit plan --root <tree> --checkpoint qwen3.6/qwen3.6-35b-a3b-fp8-bf16head --hardware gb10 --precision recipe --format plan --mode multi_seq --rows 128","met circuit plan --root <tree> --checkpoint qwen3.6/qwen3.6-35b-a3b-fp8-bf16head --hardware gb10 --precision recipe --format plan --mode verify --rows 2","met circuit plan --root <tree> --checkpoint qwen3.6/qwen3.6-35b-a3b-fp8-bf16head --hardware gb10 --precision recipe --format plan --mode verify --rows 3","met circuit plan --root <tree> --checkpoint qwen3.6/qwen3.6-35b-a3b-fp8-bf16head --hardware gb10 --precision recipe --format plan --mode verify --rows 4","met circuit plan --root <tree> --checkpoint qwen3.6/qwen3.6-35b-a3b-fp8-bf16head --hardware gb10 --precision recipe --format plan --mode draft --rows 1"],"files":[{"path":"kernels/circuits/INSTANCES.toml","sha256":"31418b33cf20e56d353d5246079b12caa97ed3bca56ec33b2b6182076d43f32c"},{"path":"recipes/qwen3.6/qwen3.6-35b-a3b-fp8-bf16head.yaml","sha256":"fdc559439b8588d100bea0de2eeaf29cba75445516e9f619b137cb68a65e207a"},{"path":"kernels/gb10/HARDWARE.toml","sha256":"24901961e415fe6a90afba954dfaa0f523f907351f42f85d9da4c19c8f7842db"},{"path":"kernels/DEVICES.toml","sha256":"256d85840f0e77ec84c5fc739f30aacff6bf24775cc8bef344e36647e51e9c25"},{"path":"kernels/circuits/qwen3_6_moe.toml","sha256":"41ef9de162e4e2b73c77e47bacdfc1c4bb3d9cb0683b255826c89f0002cb8de0"},{"path":"kernels/circuits/blocks/qwen3_hybrid.toml","sha256":"cfe66fc69723840ffe0ff9c5c12b91bb51935aef45898d5ae56567b5f2d38ab6"},{"path":"kernels/gb10/common/KERNEL_FAMILIES.toml","sha256":"4b8e793753ee539110969ff27da74aaa7d4685a612f5ba7b4062a86dc92b6c51"},{"path":"kernels/circuits/precision/qwen3.6-35b-a3b-fp8-bf16head.toml","sha256":"fc94e079613adb41d7f986f1e4d761a45cb490c87a7697240b56c7c4fd81e276"}]}}
