1. 08 Jan, 2026 1 commit
  2. 07 Jan, 2026 1 commit
    • laibao's avatar
      perf: 加速 v1 InputBatch.add_request 的 token_ids 拷贝 · 8da572a9
      laibao authored
      新增环境变量开关 VLLM_V1_FAST_TOKEN_ID_COPY(默认关闭)。开启后在 CachedRequestState 中缓存 prompt_token_ids 的 np.int32,并在 add_request 里用 np.copyto 写入 token_ids_cpu,避免长 prompt 场景反复 list->NumPy 转换开销(尤其是抢占/反复进出 batch 时)
      8da572a9
  3. 25 Sep, 2025 1 commit
  4. 19 Sep, 2025 1 commit
  5. 18 Sep, 2025 1 commit
  6. 12 Sep, 2025 1 commit
  7. 11 Sep, 2025 1 commit
  8. 08 Sep, 2025 1 commit
  9. 07 Sep, 2025 1 commit
  10. 06 Sep, 2025 1 commit
  11. 03 Sep, 2025 1 commit
  12. 02 Sep, 2025 1 commit
  13. 27 Aug, 2025 1 commit
  14. 25 Aug, 2025 1 commit
  15. 23 Aug, 2025 1 commit
  16. 21 Aug, 2025 2 commits
  17. 18 Aug, 2025 1 commit
  18. 17 Aug, 2025 1 commit
  19. 16 Aug, 2025 1 commit
  20. 13 Aug, 2025 1 commit
  21. 04 Aug, 2025 1 commit
  22. 23 Jul, 2025 1 commit
  23. 18 Jul, 2025 1 commit
  24. 02 Jul, 2025 1 commit
  25. 23 Jun, 2025 1 commit
  26. 19 Jun, 2025 1 commit
  27. 18 Jun, 2025 1 commit
  28. 10 Jun, 2025 1 commit
  29. 03 Jun, 2025 2 commits
  30. 23 May, 2025 1 commit
  31. 21 May, 2025 1 commit
  32. 15 May, 2025 1 commit
  33. 10 May, 2025 1 commit
  34. 26 Apr, 2025 1 commit
  35. 01 Apr, 2025 1 commit
  36. 28 Mar, 2025 1 commit
  37. 24 Mar, 2025 1 commit
  38. 16 Mar, 2025 1 commit