MinMoTrainingStage

Enums · AiDotNet.TextToSpeech.MultiModal

MinMo’s training stages (§3.4).

Fields

Field Summary
MinMoTrainingStage Duplex Duplex interaction alignment: the full duplex predictor alone.
MinMoTrainingStage FullAlign Speech-to-text full alignment: the input projector and the voice encoder.
MinMoTrainingStage InstructionTuning Speech-to-text instruction fine-tuning: the LLM through LoRA.
MinMoTrainingStage PreAlign Speech-to-text pre-alignment: the input projector alone, on a tenth of the data.
MinMoTrainingStage SpeechToSpeech Speech-to-speech alignment: the output projector and the voice token LM on spoken dialogues.
MinMoTrainingStage TextToSpeech Text-to-speech alignment, second part: the output projector and the voice token LM.
MinMoTrainingStage TextToSpeechProjector Text-to-speech alignment, first part: the output projector alone.
int value__