Commits

Branch selector

 main

User selector

xyang16

All time

Commit History

Commits on Apr 24, 2026

• Support only half types for concat_mla_q kernel (#37892)

  Show description for cf8a613
  xyang16

  authored
  cf8a613

  Copy full SHA for cf8a613

Commits on Apr 2, 2026

• Revert "[Kernel] Add gpt-oss Router GEMM kernel (#37205)" (#38778)

  Show description for 9bd7231
  xyang16

  authored
  9bd7231

  Copy full SHA for 9bd7231

Commits on Mar 26, 2026

• Disable dual stream execution of input projection for Qwen3 (#38152)

  Show description for 9704a5c
  xyang16

  authored
  9704a5c

  Copy full SHA for 9704a5c

Commits on Mar 20, 2026

• [Perf] Eliminate redundant SparseMatrix creation in gpt_oss_triton_kernels (#37683)

  Show description for d0532bf
  xyang16

  authored
  d0532bf

  Copy full SHA for d0532bf

Commits on Mar 18, 2026

• [Kernel] Add gpt-oss Router GEMM kernel (#37205)

  Show description for b1169d7
  xyang16

  authored
  b1169d7

  Copy full SHA for b1169d7

• [Perf] Enable dual stream execution of input projection for Qwen3 (#36795)

  Show description for f174000
  xyang16

  authored
  f174000

  Copy full SHA for f174000

Commits on Mar 9, 2026

• [Kernel] Add fused_sigmoid_gating_delta_rule_update kernel for Qwen3 Next (#35777)

  Show description for dc6b578
  xyang16

  authored
  dc6b578

  Copy full SHA for dc6b578

Commits on Mar 5, 2026

• [Perf] Optimize FusedMoEModularKernel output tensor using torch.empty (#35794)

  Show description for f917020
  xyang16

  authored
  f917020

  Copy full SHA for f917020

Commits on Feb 25, 2026

• [Bugfix] Fix expert_ids padding values in moe_align_block_size kernel (#35161)

  Show description for 3bbb204
  xyang16

  authored
  3bbb204

  Copy full SHA for 3bbb204

Commits on Feb 24, 2026

• [LoRA] Update LoRA expand kernel block_n calculation (#32621)

  Show description for c870eb9
  xyang16

  authored
  c870eb9

  Copy full SHA for c870eb9

• [Bugfix] Fix lora_ids in FusedMoE LoRA test (#35135)

  Show description for a7c9f7b
  xyang16

  authored
  a7c9f7b

  Copy full SHA for a7c9f7b

Commits on Feb 21, 2026

• [Kernel] Optimize sample_recovered_tokens_kernel (#34974)

  Show description for 7a5adad
  xyang16

  authored
  7a5adad

  Copy full SHA for 7a5adad

Commits on Feb 20, 2026

• [Kernel] Optimize grouped topk kernel (#34206)

  Show description for b1c4f0b
  xyang16

  authored
  b1c4f0b

  Copy full SHA for b1c4f0b

Commits on Feb 7, 2026

• [Bugfix] Fix _fused_moe_lora_expand signature mismatch (#33821)

  Show description for 3920caf
  xyang16

  authored
  3920caf

  Copy full SHA for 3920caf

Commits on Feb 6, 2026

• [Perf] Disable clean_logits in deepgemm fp8_mqa_logits kernel (#33568)

  xyang16

  authored
  79028d4

  Copy full SHA for 79028d4

Commits on Jan 24, 2026

• [Bugfix] Fix FusedMoE LoRA kernel offs_token out of bound value (#32279)

  Show description for ecc3dd6
  xyang16

  authored
  ecc3dd6

  Copy full SHA for ecc3dd6

Commits on Jan 23, 2026

• [Bugfix] Disable tma_aligned_scales in test_fusions_e2e (#32916)

  Show description for 90c2007
  xyang16

  authored
  90c2007

  Copy full SHA for 90c2007

Commits on Jan 22, 2026

• [Perf] Create TMA-aligned input scale tensor for DeepGemm on Hopper (#32619)

  Show description for d08b356
  xyang16

  authored
  d08b356

  Copy full SHA for d08b356

Commits on Jan 21, 2026

• [Kernel] Add topk_sigmoid kernel (#31246)

  Show description for 63227ac
  xyang16

  authored
  63227ac

  Copy full SHA for 63227ac

Commits on Jan 16, 2026

• [LoRA] Update LoRA expand kernel heuristic (#32425)

  Show description for a884bc6
  xyang16

  authored
  a884bc6

  Copy full SHA for a884bc6

Commits on Jan 10, 2026

• [LoRA][Perf] Improve FusedMoE LoRA performance for small rank (#32019)

  Show description for 543c23b
  xyang16

  authored
  543c23b

  Copy full SHA for 543c23b

Commits on Jan 9, 2026

• [Bugfix] Fix Triton FusedMoE LoRA (#30585)

  Show description for e7b68f4
  xyang16

  authored
  e7b68f4

  Copy full SHA for e7b68f4

• [Bugfix] Fix FusedMoE LoRA w2_output_size (#31949)

  Show description for 707b240
  xyang16

  authored
  707b240

  Copy full SHA for 707b240

• [Bugfix] Fix typo in FusedMoE LoRA reshape comment (#31992)

  Show description for 8413868
  xyang16

  authored
  8413868

  Copy full SHA for 8413868

Commits on Jan 7, 2026

• [Kernel] Support bias type in grouped_topk kernel (#31781)

  Show description for 0ada960
  

  xyang16

  and

  mgoin

  authored
  0ada960

  Copy full SHA for 0ada960

Commits on Dec 18, 2025

• [LoRA] Set default MXFP4 LoRA backend to Marlin (#30598)

  Show description for 9a5e965
  

  xyang16

  and

  DarkLight1337

  authored
  9a5e965

  Copy full SHA for 9a5e965

Commits on Dec 12, 2025

• [Perf] Set split_k to 1 for triton_kernels (#30528)

  Show description for 1f19d8f
  xyang16

  authored
  1f19d8f

  Copy full SHA for 1f19d8f

Commits on Nov 30, 2025

• [LoRA] Support FusedMoE LoRA Triton kernel for mxfp4 (#29708)

  Show description for a491b09
  

  xyang16

  and

  jeejeelee

  authored
  a491b09

  Copy full SHA for a491b09

Commits on Nov 28, 2025

• [LoRA] Support FusedMoE LoRA Triton kernel for mxfp4 (#28971)

  Show description for 745a3ba
  

  xyang16

  and

  jeejeelee

  authored
  745a3ba

  Copy full SHA for 745a3ba

Commits on Nov 26, 2025

• [Kernel] Use pre-allocated output buffer for triton kernel fused_experts (#29219)

  Show description for 53d7f1f
  xyang16

  authored
  53d7f1f

  Copy full SHA for 53d7f1f

Commits on Nov 19, 2025

• [Bugfix] Fix FusedMoEModularKernel for triton backend (#28913)

  Show description for 468a8d7
  xyang16

  authored
  468a8d7

  Copy full SHA for 468a8d7

Commits on Nov 11, 2025

• [Kernel] Optimize rms_norm kernel (#27931)

  Show description for 6c3c0f8
  xyang16

  authored
  6c3c0f8

  Copy full SHA for 6c3c0f8

• Fix rotary embedding benchmark script (#28323)

  Show description for 57201a6
  xyang16

  authored
  57201a6

  Copy full SHA for 57201a6

Commits on Aug 30, 2025

• Add routed_scaling_factor to MoE grouped topk (#23123)

  Show description for 8fb85b7
  

  3 peopleauthored
  8fb85b7

  Copy full SHA for 8fb85b7

Commits on Aug 25, 2025

• [Kernel] Add fused grouped_topk kernel for MoE (#23274)

  Show description for 8a3cd90
  

  xyang16

  and

  yewentao256

  authored
  8a3cd90

  Copy full SHA for 8a3cd90