ValueError: Cannot load file containing pickled data when allow_pickle=False

[kzchen95]

Getting this error when trying to train the model. Runtime error when first epoch starts. Changed the melspec = torch.from_numpy(np.load(filename)) to melspec = torch.from_numpy(np.load(filename, allow_pickle=True)) which still failed.

[KyKhaw]

Ok, this issue is related to the issue of transcripts not generating properly: https://github.com/justinjohn0306/FakeYou-Tacotron2-Notebook/issues/15

I just changed the file paths in the list.txt file from .wav to .npy and it was resolved for me.

[jooshwa]

Ok, this issue is related to the issue of transcripts not generating properly: #15

I just changed the file paths in the list.txt file from .wav to .npy and it was resolved for me.

Stuck here too. Is that all you did, just change the list.txt file? I tried that and got FileNotFoundError. I tried changing the actual .wav files as well (renamed them all .npy instead of .wav) but that gives me the same pickled data issue.

[KyKhaw]

Your .wav file should already have the .npy versions along with it. Did you run cell 6 to convert .wav files into mel spectrograms? Also be sure to change ALL filepaths in the list.txt file to .npy

[guidevops]

Your .wav file should already have the .npy versions along with it. Did you run cell 6 to convert .wav files into mel spectrograms? Also be sure to change ALL filepaths in the list.txt file to .npy

Changing the list.txt to .npy in file paths fixed the problem for me.

[guidevops]

I just discovered something: changing .npy to .wav causes the list.txt file to be saved empty. This happens because the function checks if the file referring to that transcription exists before saving the line in list.txt and the .npy file is only generated a few steps further.