kanshan/TensorRT-LLMs
1
0
Fork 0
mirror of https://github.com/NVIDIA/TensorRT-LLM.git synced 2026-01-13 22:18:36 +08:00
Code Issues Actions 1 Packages Projects Releases Wiki Activity

Resolve pipeline failures.

Browse Source 
Signed-off-by: SimengLiu-nv <simengl@nvidia.com>
This commit is contained in:
SimengLiu-nv 2026-01-06 13:34:28 -08:00
parent a4fb3f0661
commit 59261df527
1 changed files with 2 additions and 2 deletions
Show Stats Download Patch File Download Diff File Expand all files Collapse all files

4
tensorrt_llm/executor/result.py
View File

@ -625,7 +625,7 @@ class DetokenizedGenerationResultBase(GenerationResultBase):
def _handle_response(self, response: "GenerationExecutor.Response"):
# Save token lengths before processing to detect which outputs received new tokens
prev_token_lens = {o.index: len(o.token_ids) for o in self._outputs}
prev_token_lens = {id(o): len(o.token_ids) for o in self._outputs}
GenerationResultBase._handle_response(self, response)
@ -647,7 +647,7 @@ class DetokenizedGenerationResultBase(GenerationResultBase):
# to prevent re-decoding the same tokens multiple times
output_received_new_tokens = len(
beam_output.token_ids) != prev_token_lens.get(
beam_output.index, 0)
id(beam_output), 0)
if not output_received_new_tokens:
continue
if hasattr(