mstar.model.nemotron_duplex.components

Nemotron-VoiceChat components: the Nemotron-H backbone, the Fast-Conformer STT perception + RNN-T head, the EarTTS talker and the RVQ audio codec.

Modules

audio_codec	EarTTS RVQ audio codec (tts_model.audio_codec.*).
conformer	Fast-Conformer STT perception stack (stt_model.perception.*).
rnnt	RNN-T transcription head (stt_model.rnnt_decoder.* / rnnt_joint.*).