提交歷史

作者 SHA1 備註 提交日期
  AlpinDale a89c9a0e92 fix: device ordinal issues with world_size and stuff 7 月之前
  AlpinDale 427ab15434 fix: check_health when world_size==1 7 月之前
  AlpinDale 05d6e43244 fix: `torch.compile()` with mp executor backend 7 月之前
  AlpinDale 5b0c11d190 support pipeline parallel pynccl groups 7 月之前
  AlpinDale de62ceb18c refactor: eliminate parallel worker per-step task scheduling overhead 7 月之前
  AlpinDale 236be273e5 feat: tensor parallel speculative decoding (#554) 7 月之前
  AlpinDale c6a501f682 add multiprocessing executor; make ray optional 7 月之前