Method next_audio_from_audio

model.py:321–326

(self, audio_data: T.Tensor, temps=(0.8, (0.5, 0.1)))

        
    @T.no_grad()
    def next_audio_from_audio(self, audio_data: T.Tensor, temps=(0.8, (0.5, 0.1))):
        latents_in = self.tokenize(audio_data)
        next_latents = self.next_latent(latents_in, temps)
        next_model_latent = next_latents[..., self.c.latent_size:]
        audio_decoded = self.untokenize(next_model_latent)[..., -2000:]
        return audio_decoded

        
    @T.no_grad()