howard1337/S2VC

Cannot find f2114342ff9e813e18a580fa41418aee9925414e in https://github.com/s3prl/s3prl

jerrymatjila opened this issue · 2 comments

jerrymatjila commented

Running convert_batch.py throws ValueError: Cannot find f2114342ff9e813e18a580fa41418aee9925414e in https://github.com/s3prl/s3prl that originates from

S2VC/data/feature_extract.py

Line 18 in 8a6dceb

torch.hub.load("s3prl/s3prl:f2114342ff9e813e18a580fa41418aee9925414e", feature_name, refresh=True).eval().to(device) 

File "convert_batch.py", line 61, in main
src_feat_model = FeatureExtractor(src_feat_name, wav2vec_path, device)
File "/deepmind/experiments/howard1337/s2vc/data/feature_extract.py", line 18, in __init__
torch.hub.load("s3prl/s3prl:f2114342ff9e813e18a580fa41418aee9925414e", feature_name, refresh=True).eval().to(device)
File "/storage/usr/conda/envs/s2vc/lib/python3.8/site-packages/torch/hub.py", line 402, in load
repo_or_dir = _get_cache_or_reload(repo_or_dir, force_reload, verbose, skip_validation)
File "/storage/usr/conda/envs/s2vc/lib/python3.8/site-packages/torch/hub.py", line 190, in _get_cache_or_reload
_validate_not_a_forked_repo(repo_owner, repo_name, branch)
File "/storage/usr/conda/envs/s2vc/lib/python3.8/site-packages/torch/hub.py", line 160, in _validate_not_a_forked_repo
raise ValueError(f'Cannot find {branch} in https://github.com/{repo_owner}/{repo_name}. '
ValueError: Cannot find f2114342ff9e813e18a580fa41418aee9925414e in https://github.com/s3prl/s3prl. If it's a commit from a forked repo, please call hub.load() with forked repo directly.

Any idea on how to solve this?

andylida commented

Hi everyone. I figure this problem out and here are some notes.

  1. replace the 'cpc' line13 in feature_extract.py to 'modified_cpc' which has changed name in s3prl.
  2. replace 's3prl/s3prl:f2114342ff9e813e18a580fa41418aee9925414e ' by 's3prl/s3prl:v0.3.4'.
  3. replace _LinearWithBias with NonDynamicallyQuantizableLinear if you have torch >=1.8

after those finished, you can reproduce the result in that paper.
However, I used an off-the-shelf spk encoder called Wespeaker, and the EER doesn't seem good, I compare the vector with both reference and source, and the prob of misclassification is larger than 10%.Also, the WER result by Wenet is only 27%, pretty different from the results by s3prl.

Anyone is welcomed to discuss about this with me.