mrq/vall-e
5
9
Fork 0
Code Issues 8 Pull Requests Packages Projects Releases Wiki Activity

ugh

Browse Source
This commit is contained in:
mrq 2024-05-18 12:02:56 -05:00
parent 59ef9461f8
commit 74e531d391
2 changed files with 3 additions and 2 deletions
Show Stats Download Patch File Download Diff File Expand all files Collapse all files

2
scripts/process_dataset.py
View File

@ -144,7 +144,7 @@ for dataset_name in sorted(os.listdir(f'./{input_audio}/')):
i = i + 1 i = i + 1
outpath = Path(f'./{output_dataset}/{dataset_name}/{speaker_id}/{fname}_{id}.{extension}') outpath = Path(f'./{output_dataset}/{dataset_name}/{speaker_id}/{fname}_{id}.{extension}')
text = metadata[filename]["text"] text = segment["text"]
if len(text) == 0: if len(text) == 0:
continue continue

3
vall_e/emb/qnt.py
View File

@ -285,7 +285,8 @@ def encode(wav: Tensor, sr: int = cfg.sample_rate, device="cuda", levels=cfg.mod
wav = convert_audio(wav, sr, model.sample_rate, model.channels) wav = convert_audio(wav, sr, model.sample_rate, model.channels)
wav = wav.to(device) wav = wav.to(device)
encoded_frames = model.encode(wav) with torch.autocast("cuda", dtype=cfg.inference.dtype, enabled=cfg.inference.amp):
encoded_frames = model.encode(wav)
qnt = torch.cat([encoded[0] for encoded in encoded_frames], dim=-1) # (b q t) qnt = torch.cat([encoded[0] for encoded in encoded_frames], dim=-1) # (b q t)
return qnt return qnt