Closed marypilataki closed 2 years ago
Hi,
Thanks for building off of our work! Notice that this patch fixes this issue by ensuring that strict=False
when loading in a model (see this PyTorch forum thread for more info about what this does). You need to apply this patch to jukebox/jukebox/make_models.py
as done on this line of the Dockerfile.
Hope this helps.
Great, thank you very much for your response, this has resolved my issue.
Best wishes
Hello,
Thank you for making your work public. I am having an issue when trying to extract a jukebox representation using the model "5b". My python script is identical to your main under /representations/jukebox where I am using a different dataset.
Please see the exact error below