Skip to content

vllm_omni.worker.gpu_generation_worker ¶

logger `module-attribute` ¶

logger = init_logger(__name__)

GPUGenerationWorker ¶

Bases: OmniWorkerMixin, OmniGPUWorkerBase

GPU Worker for Generation model (non-autoregressive waveform generation).

Selected for stages whose pipeline topology uses execution_type=StageExecutionType.LLM_GENERATION.

model_runner_cls `class-attribute` `instance-attribute` ¶

model_runner_cls = GPUGenerationModelRunner

compile_or_warm_up_model ¶

compile_or_warm_up_model() -> CompilationTimes

Generation stages have no KV cache or sampler — skip warmup_kernels.

init_device ¶

init_device()