Skip to content

vllm_omni.model_executor.models.yue2 ¶

YuE2-3B text-to-music model package.

Modules:

Name	Description
`pipeline`	YuE2-3B pipeline topology.
`yue2`	YuE2-3B text-to-music: single-stage native-AR model, ported from upstream

YUE2_PIPELINE `module-attribute` ¶

YUE2_PIPELINE = PipelineConfig(
    model_type="yue2",
    default_deploy_config_name="yue2.yaml",
    model_arch="Yue2ForCausalLM",
    stages=(
        StagePipelineConfig(
            stage_id=0,
            model_stage="yue2",
            execution_type=StageExecutionType.LLM_AR,
            input_sources=(),
            final_output=True,
            final_output_type="audio",
            owns_tokenizer=True,
            engine_output_type="audio",
            sampling_constraints={
                "detokenize": False,
                "stop_token_ids": list(STOP_TOKEN_IDS),
            },
        ),
    ),
)