All pull requests

Pull requests

• fix: reject cached and stale benchmark outputs

  #432·chnclsr opened Oct 8, 2026·

  ·
  1

  5

• fix: handle matching infinities in evaluator

  #429·arlisgreat opened Jul 23, 2026·

  ·
  4

• Add model onboarding workflow for SGLang runtime workload collection

  #428·Rockyeast opened Jul 3, 2026·

  ·
  2

• Add setup-aware kernel-only timing for benchmark workloads

  #427·MengYu10151 opened Jul 1, 2026·

  ·
  21

• feat: add low-level GPU judge server

  #426·Ubospica opened Jun 11, 2026Collaborator·

  ·
  21

• feat(scripts): sanitize FlashInfer level-3 logs into workload JSONL

  #425·yyihuang opened May 1, 2026Contributor·4 tasks done·

  ·
  10

• feat: onboard Kimi K2.6 for mla_paged_decode_h8_ckv512_kpe64_ps1

  #410·flashinfer-bot opened Apr 21, 2026Contributor·

  ·
  6

• feat: add top_k_sampling_from_probs_v163840 definition (Kimi K2.6)

  #409·flashinfer-bot opened Apr 21, 2026Contributor·

  ·
  8

• feat: add gemm_n3072_k8192 definition

  #408·flashinfer-bot opened Apr 21, 2026Contributor·

  ·
  4

• feat: add mla_paged_decode_h8_ckv512_kpe64_ps1 definition (Kimi K2.6)

  #407·flashinfer-bot opened Apr 21, 2026Contributor·

  ·
  5

• feat: add model:kimi-k2.6 coverage for mla_paged_decode_h8_ckv512_kpe64_ps1

  #404·flashinfer-bot opened Apr 21, 2026Contributor·

  ·
  4

• feat: add gemm_n5120_k3072 and gemm_n3072_k3072 definitions (Llama 3.2 3B)

  #403·flashinfer-bot opened Apr 21, 2026Contributor·

  ·
  6

• feat: add Llama 3.2 1B model entry (coverage + sglang configs + web catalog)

  #402·ksgr5566 opened Apr 20, 2026·

  ·
  17

• feat: integrate Kimi K2.5 (model coverage + serving configs; partial, MoE blocked)

  #401·ksgr5566 opened Apr 18, 2026·

  ·
  21

• feat: add fused_add_rmsnorm_h6144 definition, workload, and coverage

  #400·ksgr5566 opened Apr 16, 2026·

  ·
  8

• feat: add rmsnorm_h6144 definition, workloads, and reference test

  #399·000FLMS opened Apr 16, 2026·

  ·
  8

• feat: add trtllm_fp8_block_scale_moe_topk8_e256_h3072_i1536 definition (MiniMax M2)

  #397·yyihuang opened Apr 14, 2026Contributor·2 tasks·

  ·
  6

• fix: improve sampling evaluator for large-nucleus distributions

  #396·yyihuang opened Apr 14, 2026Contributor·2 tasks·

  ·
  6

• feat: add gqa_paged_decode_h5_kv1_d128_ps64 definition (Llama 4 Scout/Maverick, TP=8)

  #389·yyihuang opened Apr 13, 2026Contributor·2 tasks·

  ·
  4

• feat: add moe_fp8_block_scale_ds_routing_topk8_ng1_kg1_e256_h3072_i1536 definition

  #377·yyihuang opened Apr 12, 2026Contributor·3 tasks done·

  ·
  11

• feat: add gemm_n3072_k8192 definition and reference test

  #375·yyihuang opened Apr 12, 2026Contributor·

  ·
  3

• feat: add gemm_n16384_k3072 definition and reference test

  #374·yyihuang opened Apr 12, 2026Contributor·

  ·
  7

• feat: add top_k_top_p_sampling_from_probs_v200064 reference test (MiniMax M2)

  #373·yyihuang opened Apr 12, 2026Contributor·

  ·
  3

• feat: add top_p_sampling_from_probs_v200064 reference test (MiniMax M2)

  #372·yyihuang opened Apr 12, 2026Contributor·

  ·
  5

• feat: add top_k_sampling_from_probs_v200064 reference test (MiniMax M2)

  #371·yyihuang opened Apr 12, 2026Contributor·

  ·
  3