[Optimization] compulte real max_logprobs in batch (#5430) (#5448)

This commit is contained in:
chen
2025-12-09 16:48:06 +08:00
committed by GitHub
parent 2c55bbc3f8
commit b491dcd23c
4 changed files with 48 additions and 7 deletions
@@ -53,6 +53,8 @@ class SamplingMetadata:
stop_flags: Optional[paddle.Tensor] = None
prompt_ids: Optional[paddle.Tensor] = None
prompt_lens: Optional[paddle.Tensor] = None
temp_scaled_logprobs_flag: Optional[bool] = None
top_p_normalized_logprobs_flag: Optional[bool] = None
temp_scaled_logprobs: Optional[paddle.Tensor] = None
top_p_normalized_logprobs: Optional[paddle.Tensor] = None
share_inputs: Optional[Dict[str, paddle.Tensor]] = None