[V1] Fix Detokenizer loading in AsyncLLM (#10997)

Signed-off-by: Roger Wang <[email protected]>
vllm-project · Dec 9, 2024 · c690357 · c690357
1 parent d1c2e15
commit c690357
Showing 1 changed file with 6 additions and 1 deletion.
diff --git a/vllm/v1/engine/async_llm.py b/vllm/v1/engine/async_llm.py
@@ -65,7 +65,12 @@ def __init__(
                                    input_registry)
 
         # Detokenizer (converts EngineCoreOutputs --> RequestOutput).
-        self.detokenizer = Detokenizer(vllm_config.model_config.tokenizer)
+        self.detokenizer = Detokenizer(
+            tokenizer_name=vllm_config.model_config.tokenizer,
+            tokenizer_mode=vllm_config.model_config.tokenizer_mode,
+            trust_remote_code=vllm_config.model_config.trust_remote_code,
+            revision=vllm_config.model_config.tokenizer_revision,
+        )
 
         # EngineCore (starts the engine in background process).
         self.engine_core = EngineCoreClient.make_client(