-
Zijing Liu authored
[MRv2]fix: model accuracy regression caused by reusing the stale last_sampled_tokens and draft_tokens (#39833) Signed-off-by:Zijing Liu <liuzijing2014@gmail.com>
9a6a66f3
[MRv2]fix: model accuracy regression caused by reusing the stale last_sampled_tokens and draft_tokens (#39833)
Signed-off-by:
Zijing Liu <liuzijing2014@gmail.com>