diff --git a/vllm/config/vllm.py b/vllm/config/vllm.py index 0310e8aedd..095809d543 100644 --- a/vllm/config/vllm.py +++ b/vllm/config/vllm.py @@ -728,13 +728,13 @@ class VllmConfig: "precision for chunked prefill triton kernels." ) - if ( - self.optimization_level > OptimizationLevel.O0 - and self.model_config is not None - and self.model_config.enforce_eager - ): - logger.warning("Enforce eager set, overriding optimization level to -O0") - self.optimization_level = OptimizationLevel.O0 + if self.model_config is not None and self.model_config.enforce_eager: + logger.warning( + "Enforce eager set, disabling torch.compile and CUDAGraphs. " + "This is equivalent to setting -cc.mode=none -cc.cudagraph_mode=none" + ) + self.compilation_config.mode = CompilationMode.NONE + self.compilation_config.cudagraph_mode = CUDAGraphMode.NONE if self.compilation_config.backend == "eager" or ( self.compilation_config.mode is not None