[generate] fix default autocompile case on gpu (#37756)

src/transformers/modeling_utils.py

@@ -5270,6 +5270,7 @@ class PreTrainedModel(nn.Module, ModuleUtilsMixin, PushToHubMixin, PeftAdapterMi
         # Only reset it if not present or different from previous config
         if "llama4" in self.config.model_type:  # TODO try to enable for FULL COMPILE HYBRID CACHE SUPPORT
             return self.__call__
+        compile_config = compile_config or CompileConfig()
         default_config = getattr(self.generation_config, "compile_config", None) or CompileConfig()
         if (
             not hasattr(self, "_compiled_call")