about save_musdb_XL_train_wave.py of Musdb-XL-train

[will46616]

Hi, I've followed the instruction of Musdb-XL-train website, but I got some mismatches between ratio from npy file and musdb_hq dataset. File "/home3/will/save_musdb_XL_train_wave.py", line 72, in main output = mixture * ratio ValueError: operands could not be broadcast together with shapes (2,10053632) (2,10061547)

[jeonchangbin49]

Hi, it seems that there occurs a size mismatch problem between the audio and .npy files. Could you let me know which file caused the problem? Simple print("fixed_song") after line 58 will show you that.

[will46616]

fixed_song ['The Districts - Vermont', '-5.16', '1.0'] (2, 10053632) (2, 10061547) 0%| | 0/100 [00:13<?, ?it/s] Traceback (most recent call last): File "/home3/will/save_musdb_XL_train_wave.py", line 147, in main() File "/home3/will/save_musdb_XL_train_wave.py", line 73, in main output = mixture * ratio ValueError: operands could not be broadcast together with shapes (2,10053632) (2,10061547)

Thank you for replying, the file is 'The Districts - Vermont'.

[jeonchangbin49]

Hmm, it's weird. I think it's very unlikely that your 'The Districts - Vermont' and my 'The Districts - Vermont' have different lengths because it seems that there is only one version of musdb18-hq available in https://zenodo.org/records/3338373. I'll get back to you after I check the audio length maybe this weekend or next week. Let me know if you have more info regarding the problem.

[will46616]

Thanks for your reply, I downloaded my MUSDB18-HQ dataset from the same URL. I will inform you if there is any other information.