Skip to content

vllm_omni.model_executor.models.cosyvoice3

Modules:

Name Description
code2wav_core
cosyvoice3
cosyvoice3_code2wav

CosyVoice3 Code2Wav Stage - Converts speech tokens to audio waveforms.

cosyvoice3_talker
flow_estimator_trt

TensorRT engine for the CosyVoice3 flow-decoder (CFM) DiT estimator.

pipeline

CosyVoice3 pipeline topology (frozen).

runtime

Runtime toggles shared by the CosyVoice3 stages.

speaker_embedding_trt

TensorRT speaker-embedding (campplus) extractor for CosyVoice3.

tokenizer
utils