vllm_omni.entrypoints.cli.benchmark.serve ¶

Online serving benchmark subcommand for vLLM-Omni.

OmniBenchmarkServingSubcommand starts from vLLM's serving benchmark arguments, adds Omni-specific options and adaptations, then preprocesses the parsed namespace and delegates execution to vllm_omni.benchmarks.serve. This keeps CLI integration separate from the multimodal benchmark runtime.

help = "Benchmark online serving. Supports Daily-Omni, OmniInteract, and Seed-TTS datasets."

name = 'serve'

add_cli_args(parser: ArgumentParser) -> None

cmd(args: Namespace) -> None