vllm_omni.model_executor.models.glm_tts.pipeline ¶

GLM_TTS_PIPELINE = PipelineConfig(
    model_type="glm_tts",
    model_arch="GLMTTSForConditionalGeneration",
    stages=(
        StagePipelineConfig(
            stage_id=0,
            model_stage="glm_tts",
            execution_type=LLM_AR,
            input_sources=(),
            owns_tokenizer=True,
            engine_output_type="latent",
            async_chunk_process_next_stage_input_func=f"{_PROC}.ar_to_dit_async_chunk",
            sampling_constraints={
                "stop_token_ids": [59253]
            },
        ),
        StagePipelineConfig(
            stage_id=1,
            model_stage="glm_tts_dit",
            execution_type=LLM_GENERATION,
            input_sources=(0,),
            final_output=True,
            final_output_type="audio",
            engine_output_type="latent",
            sync_process_input_func=f"{_PROC}.ar_to_dit",
        ),
    ),
)