1. 06 Feb, 2026 1 commit
  2. 23 Jan, 2026 1 commit
  3. 21 Jan, 2026 1 commit
    • zhuwenwen's avatar
      perf(qwen3): 融合 q/k RMSNorm + RoPE · 3cd3d1e6
      zhuwenwen authored
      新增 VLLM_USE_FUSED_RMS_ROPE 分支,走 fused 路径
      注册 torch.ops.vllm.rms_rotary_embedding_fuse(direct_register_custom_op)
      cos_sin_cache 自动转 device/dtype 并缓存,避免每次重复拷贝
      3cd3d1e6
  4. 20 Jan, 2026 1 commit
  5. 19 Jan, 2026 1 commit
  6. 14 Jan, 2026 1 commit
  7. 07 Jan, 2026 2 commits
  8. 06 Jan, 2026 1 commit
  9. 05 Jan, 2026 4 commits
  10. 04 Jan, 2026 1 commit
  11. 22 Dec, 2025 1 commit
  12. 20 Dec, 2025 1 commit
  13. 16 Dec, 2025 1 commit
  14. 12 Dec, 2025 2 commits
  15. 30 Nov, 2025 1 commit
  16. 20 Nov, 2025 1 commit
  17. 17 Nov, 2025 1 commit
  18. 13 Nov, 2025 3 commits
  19. 31 Oct, 2025 1 commit
  20. 29 Oct, 2025 1 commit
  21. 28 Oct, 2025 1 commit
  22. 25 Oct, 2025 1 commit
  23. 21 Oct, 2025 1 commit
  24. 11 Oct, 2025 1 commit
  25. 02 Oct, 2025 1 commit
  26. 01 Oct, 2025 4 commits
  27. 29 Sep, 2025 1 commit
  28. 28 Sep, 2025 3 commits