Bhargav v
Bhargav v
I can confirm this on 1.0.20v on Fedora workstation 33 (btrfs) on both flatpak and rpm.
In transfer learning as fixed feature extractor, all parameters are counted as trainable parameters.
Thanks for the reply. It was a vgg16 model that I set `requires_grad=False` for most of the initial layers. I didn't use checkpoint. The model worked as intended but torch...
Can't believe this still isn't fixed. Facing same issue on Fedora 33
Hi, Looks like they have changed it quite a bit. 1.Have you tested your patch in colab? I tried testing it but gdrive gets disconnected after a while and gives...
Have you used any plots just before the problem starts to appear? If yes what kinds of plots? If you are comfortable with it, could you share the notebook?
Hi! I ran it on colab after passing the exact file name with underscores. It gave me the pdf file for all content as expected. Could you maybe share your...
Interesting! The pdf usually gets saved in top directory of Google drive. The api itself has changed a bit now. #3 tries to fix. It is interesting that `.ipynb` file...
This would enable us to use newer OpenAI embedding models, that support upto (3072 dims). The newer models seems to outperform the older ada002 embedding model, even with compressed to...