0 Releases 6 Tags

RSS Feed

• v0.0.3 ebdc6fed03

  Compare

  fix: pass lm_head to LogitsProcessor instead of calling forward()

  Chranos released this 2026-02-06 14:21:14 +08:00 | 56 commits to main since this release

  In vLLM v0.6.2, ParallelLMHead.forward() raises RuntimeError since
  its weights should be used through LogitsProcessor.linear_method.apply().
  Pass lm_head as first arg to LogitsProcessor which handles the
  hidden_states -> logits projection internally.

  Downloads

  • Source Code (ZIP)
  • Source Code (TAR.GZ)