Pre-commit

Signed-off-by: Reagan <reaganjlee@gmail.com>

docs/mkdocs/hooks/generate_argparse.py

@@ -223,7 +223,9 @@ def on_startup(command: Literal["build", "gh-deploy", "serve"], dirty: bool):
         "run-batch": create_parser(openai_run_batch.make_arg_parser),
         # Benchmark CLI
         "bench_latency": create_parser(bench_latency.add_cli_args),
-        "bench_multimodal_processor": create_parser(bench_multimodal_processor.add_cli_args),
+        "bench_multimodal_processor": create_parser(
+            bench_multimodal_processor.add_cli_args
+        ),
         "bench_serve": create_parser(bench_serve.add_cli_args),
         "bench_sweep_plot": create_parser(bench_sweep_plot.add_cli_args),
         "bench_sweep_plot_pareto": create_parser(bench_sweep_plot_pareto.add_cli_args),

vllm/benchmarks/multimodal_processor.py

@@ -86,7 +86,8 @@ def collect_mm_processor_stats(
 
     if debug and not any(stats_by_stage.values()):
         print(
-            "Warning: No MM processor stats found. Ensure --enable-mm-processor-stats is set."
+            "Warning: No MM processor stats found. "
+            "Ensure --enable-mm-processor-stats is set."
         )
 
     return stats_by_stage
@@ -136,28 +137,27 @@ def generate_random_multimodal_prompts(
 ) -> tuple[list[list[dict]], list[int]]:
     """
     Generate random multimodal prompts with synthetic images and text tokens.
-    
+
     Returns:
         tuple: (prompts, expected_output_lens)
             - prompts: List of OpenAI chat format messages with text and images
             - expected_output_lens: List of expected output lengths
     """
     from PIL import Image
+
     from vllm.benchmarks.datasets import process_image
-    
+
     rng = np.random.default_rng(seed)
-    
+
     prompts = []
     expected_output_lens = []
-    
+
     for i in range(num_prompts):
         vocab_size = tokenizer.vocab_size
-        prompt_token_ids = rng.integers(
+        prompt_token_ids = rng.integers(0, vocab_size, size=input_len).tolist()
-            0, vocab_size, size=input_len
+
-        ).tolist()
-        
         text_prompt = tokenizer.decode(prompt_token_ids)
-        
+
         mm_items = []
         for _ in range(num_images):
             # Generate random RGB image
@@ -168,13 +168,13 @@ def generate_random_multimodal_prompts(
             # Process to OpenAI format
             mm_item = process_image(image)
             mm_items.append(mm_item)
-        
+
         # Create chat format: text + images
         content = [{"type": "text", "text": text_prompt}]
         content.extend(mm_items)
         prompts.append([{"role": "user", "content": content}])
         expected_output_lens.append(output_len)
-    
+
     return prompts, expected_output_lens
 
 
@@ -382,7 +382,6 @@ def add_cli_args(parser: argparse.ArgumentParser) -> None:
 
 def main(args: argparse.Namespace) -> None:
     """Main entry point for the multimodal processor benchmark."""
-    from datetime import datetime
 
     print("Starting multimodal processor benchmark...")
     result = benchmark_multimodal_processor(args)

vllm/entrypoints/cli/benchmark/multimodal_processor.py

@@ -19,4 +19,3 @@ class BenchmarkMultimodalProcessorSubcommand(BenchmarkSubcommandBase):
     @staticmethod
     def cmd(args: argparse.Namespace) -> None:
         main(args)
-

vllm/multimodal/processing.py

@@ -25,7 +25,6 @@ import regex as re
 import torch
 from typing_extensions import TypeVar, assert_never
 
-import vllm.envs as envs
 from vllm.logger import init_logger
 from vllm.tokenizers import TokenizerLike
 from vllm.transformers_utils.processor import cached_processor_from_config
@@ -126,7 +125,9 @@ class MultiModalProcessorTimingStats:
         }
 
 
-def get_timing_stats_from_engine_client(engine_client: Any) -> dict[str, dict[str, float]]:
+def get_timing_stats_from_engine_client(
+    engine_client: Any,
+) -> dict[str, dict[str, float]]:
     """
     Get all timing stats from the context associated with the engine client.
 
@@ -2334,11 +2335,6 @@ class BaseMultiModalProcessor(ABC, Generic[_I]):
         if request_id is not None:
             self.info.ctx.create_timing_stats(request_id)
 
-        stats = (
-            self.info.ctx.get_timing_stats(request_id)
-            if request_id is not None
-            else None
-        )
         mm_items = self._to_mm_items(mm_data)
 
         if tokenization_kwargs is None: