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

Transit s2s clips back to CPU memory after processing

Browse Source
This commit is contained in:
James Betker 2021-12-29 08:54:07 -07:00
parent af6d5cd526
commit 8a02ba5935
1 changed files with 1 additions and 1 deletions
Show Stats Download Patch File Download Diff File Expand all files Collapse all files

2
codes/scripts/audio/speech_to_speech_clip.py
View File

@ -66,7 +66,7 @@ def process_subdir(subdir, options, clip_sz):
stacked = torch.stack(clips[:256], dim=0).cuda() stacked = torch.stack(clips[:256], dim=0).cuda()
clips = clips[256:] clips = clips[256:]
mels = wav_to_mel(stacked) mels = wav_to_mel(stacked)
outp = clip_model.inference(mels) outp = clip_model.inference(mels).cpu()
if sims is None: if sims is None:
sims = outp sims = outp
else: else: