xinyun/vllm
1
0
Fork 0
mirror of https://git.datalinker.icu/vllm-project/vllm.git synced 2026-09-09 04:07:02 +08:00
Code Issues Packages Projects Releases Wiki Activity

[Bugfix] skip cuda graph for drafter when running with eager (#26821)

Browse Source 
Signed-off-by: Benjamin Chislett <bchislett@nvidia.com>
This commit is contained in:
Benjamin Chislett 2025-10-21 18:39:09 -04:00 committed by GitHub
parent 4a8a567e16
commit 19748806f0
No known key found for this signature in database
GPG Key ID: B5690EEEBB952194
1 changed files with 4 additions and 1 deletions
Show Stats Download Patch File Download Diff File Expand all files Collapse all files

5
vllm/v1/worker/gpu_model_runner.py
View File

@ -3482,7 +3482,10 @@ class GPUModelRunner(LoRAModelRunnerMixin, KVConnectorModelRunnerMixin):
if self.speculative_config and self.speculative_config.use_eagle(): if self.speculative_config and self.speculative_config.use_eagle():
assert isinstance(self.drafter, EagleProposer) assert isinstance(self.drafter, EagleProposer)
use_cudagraphs = cudagraph_runtime_mode == CUDAGraphMode.PIECEWISE use_cudagraphs = (
cudagraph_runtime_mode == CUDAGraphMode.PIECEWISE
and not self.speculative_config.enforce_eager
)
self.drafter.dummy_run(num_tokens, use_cudagraphs=use_cudagraphs) self.drafter.dummy_run(num_tokens, use_cudagraphs=use_cudagraphs)
# This is necessary to avoid blocking DP. # This is necessary to avoid blocking DP.