Skip to content

vllm_omni.model_executor.models.voxtral_tts.pipeline

Voxtral TTS pipeline topology (frozen).

Stage 0: audio_generation — text → acoustic latents (LLM_AR, tokenizer owner). Stage 1: audio_tokenizer — acoustic latents → waveform (LLM_GENERATION).

VOXTRAL_TTS_PIPELINE module-attribute 

VOXTRAL_TTS_PIPELINE = PipelineConfig(
    model_type="voxtral_tts",
    model_arch="VoxtralTTSForConditionalGeneration",
    stages=(
        StagePipelineConfig(
            stage_id=0,
            model_stage="audio_generation",
            execution_type=LLM_AR,
            input_sources=(),
            final_output=False,
            final_output_type="text",
            owns_tokenizer=True,
            engine_output_type="latent",
            async_chunk_process_next_stage_input_func=f"{_PROC}.generator2tokenizer_async_chunk",
            sampling_constraints={"detokenize": True},
        ),
        StagePipelineConfig(
            stage_id=1,
            model_stage="audio_tokenizer",
            execution_type=LLM_GENERATION,
            input_sources=(0,),
            final_output=True,
            final_output_type="audio",
            engine_output_type="audio",
            sampling_constraints={"detokenize": True},
            extras={
                "tts_args": {"max_instructions_length": 500}
            },
        ),
    ),
)