mrq/DL-Art-School
1
3
Fork 5
Code Issues Pull Requests Packages Projects Releases Wiki Activity

Allow saving mel strips as files from the dataset implementation

Browse Source
This commit is contained in:
James Betker 2021-08-13 22:46:41 -06:00
parent cfd284f425
commit 72622b4d61
1 changed files with 14 additions and 6 deletions
Show Stats Download Patch File Download Diff File Expand all files Collapse all files

20
codes/data/audio/nv_tacotron_dataset.py
View File

@ -190,23 +190,31 @@ class TextMelCollate():
}
def save_mel_buffer_to_file(mel, path):
np.save(path, mel.numpy())
def load_mel_buffer_from_file(path):
return torch.tensor(np.load(path))
def dump_mels_to_disk():
params = {
'mode': 'nv_tacotron',
'path': 'E:\\audio\\MozillaCommonVoice\\en\\test.tsv',
'phase': 'train',
'n_workers': 0,
'batch_size': 32,
'batch_size': 1,
'fetcher_mode': 'mozilla_cv',
'needs_collate': False,
'max_mel_length': 255800,
'needs_collate': True,
'max_mel_length': 1000,
'max_text_length': 200,
'return_wavs': True,
#'return_wavs': True,
#'input_sample_rate': 22050,
#'sampling_rate': 8000
}
output_path = 'D:\\mozcv_mels'
output_path = 'D:\\dlas\\results\\mozcv_mels'
os.makedirs(os.path.join(output_path, 'clips'), exist_ok=True)
from data import create_dataset, create_dataloader
ds, c = create_dataset(params, return_collate=True)
dl = create_dataloader(ds, params, collate_fn=c)
@ -214,7 +222,7 @@ def dump_mels_to_disk():
mels = b['padded_mel']
fnames = b['filenames']
for j, fname in enumerate(fnames):
torch.save(mels[j], f'{os.path.join(output_path, fname)}_mel.pth')
save_mel_buffer_to_file(mels[j], f'{os.path.join(output_path, fname)}_mel.npy')
if __name__ == '__main__':