docs: note that enforce_eager also disables torch.compile (#55271)
Signed-off-by: Anshul Desai <36732940+AnshulDesai@users.noreply.github.com>
Anshul Desai committed
Sep 3, 2026 at 21:25 UTC
da55494a125979ffc9d4c1f17d593175cb076bf4
1 file changed
+4
-1
vllm/config/model.py
+4
-1
@@ -242,7 +242,10 @@ class ModelConfig:
242
"""Whether to always use eager-mode PyTorch. If True, we will disable CUDA
243
graph and always execute the model in eager mode. If False, we will use
244
CUDA graph and eager execution in hybrid for maximal performance and
245
- flexibility."""
245
+ flexibility.
246
+
247
+ NOTE: This disables both `torch.compile` and CUDA graphs, and is
248
+ equivalent to setting `-cc.mode=none -cc.cudagraph_mode=none`."""
249
enable_return_routed_experts: bool = False
250
"""Whether to return routed experts."""
251
return_sampling_mask: bool = False