xinyun/vllm
1
0
Fork 0
mirror of https://git.datalinker.icu/vllm-project/vllm.git synced 2025-12-09 23:45:54 +08:00
Code Issues Packages Projects Releases Wiki Activity

[Core] Force PIECEWISE CUDAGraph mode for encoder-decoder (#25701)

Browse Source 
Signed-off-by: Russell Bryant <rbryant@redhat.com>
This commit is contained in:
Russell Bryant 2025-09-25 21:21:56 -04:00 committed by GitHub
parent 53a30845be
commit 13dd93c667
No known key found for this signature in database
GPG Key ID: B5690EEEBB952194
1 changed files with 4 additions and 2 deletions
Show Stats Download Patch File Download Diff File Expand all files Collapse all files

6
vllm/config/__init__.py
View File

@ -364,9 +364,11 @@ class VllmConfig:
self.compilation_config.cudagraph_mode = \ self.compilation_config.cudagraph_mode = \
CUDAGraphMode.FULL_AND_PIECEWISE CUDAGraphMode.FULL_AND_PIECEWISE
# pooling model does not support full cudagraphs # pooling models and encoder-decoder models
# do not support full cudagraphs
if self.model_config is not None and \ if self.model_config is not None and \
self.model_config.pooler_config is not None: (self.model_config.pooler_config is not None
or self.model_config.is_encoder_decoder):
self.compilation_config.cudagraph_mode = \ self.compilation_config.cudagraph_mode = \
CUDAGraphMode.PIECEWISE CUDAGraphMode.PIECEWISE
else: else: