vLLM-Omni extensions for the vllm bench serve CLI.
Core functions
add_omni_args: Register all Omni-specific argument groups by calling the feature-specific add_*_cli_args helpers in this module. extend_omni_choices: Add Omni datasets and backends to choices defined by the upstream vLLM parser, including its shadow parser. update_omni_help: Extend upstream help text with Omni-specific behavior. preprocess_serve_args: Apply transformations that require parsed values before the serving benchmark starts.
OmniBenchmarkServingSubcommand.add_cli_args invokes the first three after upstream vLLM registers its arguments. New feature arguments should normally be added to the corresponding add_*_cli_args helper, or to a new helper called by add_omni_args.
add_daily_omni_cli_args(parser: ArgumentParser) -> None
Add CLI arguments specific to the Daily-Omni dataset.
add_diffusion_cli_args(parser: ArgumentParser) -> None
Add CLI arguments for diffusion model benchmarks.
add_multi_stage_cli_args(parser: ArgumentParser) -> None
Add CLI arguments for vLLM-Omni multi-stage benchmarks.
add_omni_args(parser: ArgumentParser) -> None
Register all vLLM-Omni serving benchmark arguments.
add_seed_tts_cli_args(parser: ArgumentParser) -> None
Add CLI arguments for Seed-TTS benchmarks.
add_videomme_cli_args(parser: ArgumentParser) -> None
Add CLI arguments specific to the Video-MME dataset.
extend_omni_choices(parser: ArgumentParser) -> None
Extend upstream argument choices with Omni-specific values.
preprocess_serve_args(args: Namespace) -> None
Apply serving benchmark CLI transformations after parsing.
update_omni_help(parser: ArgumentParser) -> None
Update upstream argument help text to describe Omni-specific behavior.