Skip to content

GitLab

Switch branch/tag
  1. 10 Feb, 2026 1 commit
    • zhuwenwen's avatar
      support fa kvcache fp8, add VLLM_USE_QUERY_QUANT to not use q quant(todo) · 0dfb30d5
      zhuwenwen authored 
      [opt] 优化epsp代码, 零消耗添加epsp
update VLLM_USE_FUSED_RMS_ROPE=0 (default). for qwen3, VLLM_USE_FUSED_RMS_ROPE=1 (default)
feat(moe/marlin): Marlin W16A16 MoE 自动探测并预打包（去掉手动开关）
perf(qwen3): 融合 q/k RMSNorm + RoPE
fused_moe_fp8接入lmslim
      0dfb30d5
  3. 04 Dec, 2025 1 commit
  5. 02 Dec, 2025 1 commit
  7. 12 Nov, 2025 1 commit
  9. 09 Nov, 2025 1 commit
  11. 07 Nov, 2025 1 commit
  13. 04 Nov, 2025 1 commit
  15. 03 Nov, 2025 1 commit
  17. 31 Oct, 2025 2 commits
  19. 24 Oct, 2025 1 commit
    • zhuwenwen's avatar
      add VLLM_USE_LIGHTOP_MOE_SUM_MUL_ADD · c2e6f453
      zhuwenwen authored 
      support prefix cache on kme
fix the error in test_moe caused by moe align not supporting 511 and 211
multi-modal switching to torch implementation on z100l&k100
      c2e6f453
  21. 23 Oct, 2025 1 commit
    • zhuwenwen's avatar
      support prefix cache on kme · 6f1db287
      zhuwenwen authored 
      fix the error in test_moe caused by moe align not supporting 511
multi-modal switching to torch implementation on z100l&k100
      6f1db287
  23. 13 Oct, 2025 1 commit
  25. 28 Sep, 2025 1 commit
  27. 21 Aug, 2025 1 commit
  29. 11 Aug, 2025 1 commit
  31. 09 Aug, 2025 1 commit
  33. 07 Aug, 2025 1 commit
  35. 29 Jul, 2025 1 commit
  37. 28 Jul, 2025 1 commit
  39. 04 Jul, 2025 1 commit
  41. 03 Jul, 2025 1 commit
  43. 02 Jul, 2025 1 commit