1. 10 Feb, 2026 1 commit
    • zhuwenwen's avatar
      [feat] 适配block和channel fp8, 添加dp attention功能 · 2b7b1a31
      zhuwenwen authored
      [pd] 修改非堆成切分的判断, pd分离使用默认调度,set VLLM_USE_PD_SPLIT=1
      [perf](fused-moe): 预打包 Marlin W16A16 MoE 权重,降低 warmup 显存峰值
      [fix]修复丢弃MTP代码报错,pp+chunkprefill多并发input ids更新bug, 修复不开启融合图的断言错误, PP 场景 decode 阶段 token 被误丢弃导致卡住
      2b7b1a31
  2. 04 Jul, 2025 2 commits
  3. 03 Jul, 2025 2 commits
  4. 02 Jul, 2025 1 commit
  5. 01 Jul, 2025 2 commits
  6. 28 Jun, 2025 1 commit
  7. 26 Jun, 2025 1 commit
  8. 11 Jun, 2025 1 commit
  9. 07 Jun, 2025 1 commit
  10. 04 Jun, 2025 1 commit
  11. 03 Jun, 2025 2 commits
  12. 27 May, 2025 1 commit
  13. 25 May, 2025 1 commit
  14. 23 May, 2025 1 commit
  15. 14 May, 2025 4 commits
  16. 13 May, 2025 1 commit
  17. 05 May, 2025 1 commit
  18. 30 Apr, 2025 1 commit
  19. 25 Apr, 2025 1 commit
  20. 22 Apr, 2025 1 commit
  21. 08 Apr, 2025 2 commits
  22. 07 Apr, 2025 1 commit
  23. 06 Apr, 2025 1 commit
  24. 01 Apr, 2025 1 commit
  25. 28 Mar, 2025 1 commit
  26. 27 Mar, 2025 1 commit
  27. 26 Mar, 2025 1 commit
  28. 24 Mar, 2025 1 commit
  29. 11 Mar, 2025 1 commit
  30. 07 Mar, 2025 1 commit
  31. 26 Feb, 2025 2 commits