Closed YIN95 closed 1 year ago
This is because the jukemirlib automatically downloads the jukebox model (which is about 10GB) into the '/usr/' directory or something like that (you can see this line). I believe the disk quota of this directory is limited on your cluster, so I think you may want to change the cache_dir to the directory of your project for example.
Actually, I could see the model can be downloaded successfully, but still has this problem (
That's weird, have you tried creating data by extracting only the baseline feature "--extract-baseline" but not "--extract-jukebox"?
"--extract-baseline" works for me! But "--extract-jukebox" encounters this issue.
Hi! Thanks for sharing the impressive work. Could I ask how to solve this? I am processing the dataset by one A100, on a cluster. But it shows disk quota is exceeded, even if it is not true.
Thank you in advance.