Return more infos for computing average acceptance length (#3152)

This commit is contained in:
Lianmin Zheng
2025-01-26 04:51:54 -08:00
committed by GitHub
parent 7e0976133c
commit 1dda8c5e4c
10 changed files with 97 additions and 15 deletions
@@ -785,6 +785,9 @@ class TokenizerManager:
i,
)
if self.server_args.speculative_algorithm:
meta_info["spec_verify_ct"] = recv_obj.spec_verify_ct[i]
if not isinstance(recv_obj, BatchEmbeddingOut):
meta_info.update(
{
@@ -809,6 +812,7 @@ class TokenizerManager:
"embedding": recv_obj.embeddings[i],
"meta_info": meta_info,
}
state.out_list.append(out_dict)
state.finished = recv_obj.finished_reasons[i] is not None
state.event.set()