vllm_omni.model_executor.models.qwen3_omni.quantization ¶

Quantization mapping helpers for nested Qwen3-Omni stages.

Qwen3OmniNestedSupportsQuant ¶

Bases: SupportsQuant

SupportsQuant variant for models initialized inside the stage wrapper.

apply_nested_quant_config_mapping(
    model: SupportsQuant,
) -> None

Avoid applying a stage weight mapper twice to a mapped config.

apply_outer_quant_config_mapping(
    model: SupportsQuant,
) -> None

Map checkpoint paths once at the unified model boundary.