Fix linter

Signed-off-by: Jeffrey Wang <jeffreywang@anyscale.com>

Author: jeffreywang-anyscale

python/ray/llm/_internal/batch/stages/vllm_engine_stage.py

@@ -506,7 +506,9 @@ async def _generate_async(self, request: vLLMEngineRequest) -> Any:
         async for request_output in stream:
             if request_output.finished:
                 # Bypass the original full prompt.
-                request_output.prompt = request.prompt if request.prompt is not None else ""
+                request_output.prompt = (
+                    request.prompt if request.prompt is not None else ""
+                )
                 return request_output
 
         raise RuntimeError(
@@ -635,7 +637,10 @@ def validate_inputs(self, inputs: List[Dict[str, Any]]):
                 )
 
         original_expected_keys = self.expected_input_keys.copy()
-        self.expected_input_keys = self.expected_input_keys - {"prompt", "tokenized_prompt"}
+        self.expected_input_keys = self.expected_input_keys - {
+            "prompt",
+            "tokenized_prompt",
+        }
 
         try:
             super().validate_inputs(inputs)

python/ray/llm/tests/batch/gpu/processor/test_vllm_engine_proc.py

@@ -203,6 +203,7 @@ def test_generation_model(gpu_type, model_opt_125m, backend):
 
 def test_generation_model_tokenized_prompt(gpu_type, model_opt_125m):
     from transformers import AutoTokenizer
+
     tokenizer = AutoTokenizer.from_pretrained(model_opt_125m, trust_remote_code=True)
 
     processor_config = vLLMEngineProcessorConfig(
@@ -293,7 +294,10 @@ def test_generation_model_missing_prompt_and_tokenized_prompt(gpu_type, model_op
         ds = ds.materialize()
 
     error_str = str(exc_info.value)
-    assert "Either 'prompt' (text) or 'tokenized_prompt' (tokens) must be provided" in error_str
+    assert (
+        "Either 'prompt' (text) or 'tokenized_prompt' (tokens) must be provided"
+        in error_str
+    )
 
 
 def test_embedding_model(gpu_type, model_smolvlm_256m):