diff --git a/vllm/engine/llm_engine.py b/vllm/engine/llm_engine.py
index 784c0caf..83dfbc2e 100644
--- a/vllm/engine/llm_engine.py
+++ b/vllm/engine/llm_engine.py
@@ -153,7 +153,7 @@ class LLMEngine:
                     placement_group=placement_group,
                     placement_group_capture_child_tasks=True),
                 **ray_remote_kwargs,
-            )(RayWorker).remote()
+            )(RayWorker).remote(self.model_config.trust_remote_code)
             self.workers.append(worker)
 
         # Initialize torch distributed process group for the workers.
diff --git a/vllm/engine/ray_utils.py b/vllm/engine/ray_utils.py
index f085f922..2982f0f9 100644
--- a/vllm/engine/ray_utils.py
+++ b/vllm/engine/ray_utils.py
@@ -11,7 +11,11 @@ try:
         """Ray wrapper for vllm.worker.Worker, allowing Worker to be
         lazliy initialized after Ray sets CUDA_VISIBLE_DEVICES."""
 
-        def __init__(self) -> None:
+        def __init__(self, init_cached_hf_modules=False) -> None:
+            if init_cached_hf_modules:
+                # pylint: disable=import-outside-toplevel
+                from transformers.dynamic_module_utils import init_hf_modules
+                init_hf_modules()
             self.worker = None
 
         def init_worker(self, worker_init_fn):