Skip to content

vllm_omni.model_executor.models.minimax_h3.checkpoint ¶

MiniMax H3 checkpoint resolution for the AR text-encoder stage.

MINIMAX_H3_ENCODER_DOWNLOAD_PATTERNS `module-attribute` ¶

MINIMAX_H3_ENCODER_DOWNLOAD_PATTERNS = {
    partition: [
        f"{subdir}/text_encoder/**",
        f"{subdir}/video_vae/**",
        f"{subdir}/audio_vae/**",
    ]
    for partition, subdir in {
        "fl2va": "FL2VA",
        "ref2va": "Ref2VA",
    }.items()
}

resolve_minimax_h3_model_root `module-attribute` ¶

resolve_minimax_h3_model_root = (
    resolve_minimax_h3_encoder_model_root
)

is_minimax_h3_modular ¶

is_minimax_h3_modular(
    model: str, revision: str | None = None
) -> bool

resolve_minimax_h3_encoder_model_root ¶

resolve_minimax_h3_encoder_model_root(
    model: str, revision: str | None, task_type: str | None
) -> str

resolve_minimax_h3_partition ¶

resolve_minimax_h3_partition(
    model: str,
    task_type: str | None,
    *,
    auto_partition: str,
) -> str