MCPcopy Create free account
hub / github.com/MoonInTheRiver/DiffSinger / build_tts_model

Method build_tts_model

usr/task.py:16–24  ·  view source on GitHub ↗
(self)

Source from the content-addressed store, hash-verified

14
15class DiffFsTask(FastSpeech2Task):
16 def build_tts_model(self):
17 mel_bins = hparams['audio_num_mel_bins']
18 self.model = GaussianDiffusion(
19 phone_encoder=self.phone_encoder,
20 out_dims=mel_bins, denoise_fn=DIFF_DECODERS[hparams['diff_decoder_type']](hparams),
21 timesteps=hparams['timesteps'],
22 loss_type=hparams['diff_loss_type'],
23 spec_min=hparams['spec_min'], spec_max=hparams['spec_max'],
24 )
25
26 def run_model(self, model, sample, return_output=False, infer=False):
27 txt_tokens = sample['txt_tokens'] # [B, T_t]

Callers

nothing calls this directly

Calls 1

GaussianDiffusionClass · 0.50

Tested by

no test coverage detected