1. 24 Dec, 2025 1 commit
    • laibao's avatar
      V1 采样器:新增 reduced top-k/top-p 采样路径 · 9b1e03d4
      laibao authored
      新增环境变量 VLLM_V1_USE_REDUCED_TOPK_TOPP_SAMPLER 用于开关控制
      扩展 SamplingMetadata,增加 max_top_k 与 has_any_no_top_k
      在 InputBatch 侧计算 top-k 的主机端汇总信息,避免 device 同步
      更新 Sampler/TopKTopPSampler 传递并使用新参数以启用优化采样
      9b1e03d4
  2. 09 Dec, 2025 1 commit
  3. 08 Dec, 2025 2 commits
  4. 05 Dec, 2025 1 commit
    • laibao's avatar
      [bugfix] 优化 reject-sampling 的 InputBatch 元数据处理 · a0d556fe
      laibao authored
      - 在 InputBatch.refresh_metadata 中为展开后的采样元数据引入 repeat_count 记录重复次数
      - 完善元数据刷新逻辑以适配 reject-sampling 优化路径
      - 更新 GPUModelRunnerBase,在 batch 处理阶段正确消费新的采样元数据与重复计数
      a0d556fe
  5. 27 Nov, 2025 1 commit
  6. 26 Nov, 2025 2 commits
  7. 21 Nov, 2025 1 commit
  8. 17 Oct, 2025 1 commit
  9. 13 Oct, 2025 1 commit
  10. 11 Oct, 2025 2 commits
  11. 10 Oct, 2025 3 commits
  12. 09 Oct, 2025 1 commit
  13. 26 Sep, 2025 1 commit
  14. 25 Sep, 2025 1 commit
  15. 24 Sep, 2025 2 commits
  16. 16 Sep, 2025 1 commit
  17. 06 Sep, 2025 1 commit
  18. 04 Sep, 2025 2 commits
  19. 03 Sep, 2025 1 commit
  20. 01 Sep, 2025 1 commit
  21. 27 Aug, 2025 1 commit
  22. 19 Aug, 2025 1 commit
  23. 18 Aug, 2025 1 commit
  24. 16 Aug, 2025 1 commit
  25. 15 Aug, 2025 1 commit
  26. 14 Aug, 2025 1 commit
  27. 13 Aug, 2025 2 commits
  28. 11 Aug, 2025 1 commit
  29. 08 Aug, 2025 1 commit
  30. 07 Aug, 2025 2 commits
  31. 06 Aug, 2025 1 commit