Closed aalbersk closed 1 year ago
@aalbersk Hi Anna, the notebook was actually made on an earlier version of DALLE-pytorch. The mask has since been removed
Let me know if https://github.com/lucidrains/DALLE-pytorch/commit/daf30d02a42d4e407b540a568950cdf085f5d9b1 corrects the issue!
Hey, yes, training works now.
Btw, to make a whole example work correctly, I needed to move all_image_code to cpu with all_image_codes = all_image_codes.cpu()
before calculating accuracy at the end.
But now everything works. Thanks for the quick fix!
@aalbersk good to hear, and cute corgi!
Hi, I received similar issue to one reported in https://github.com/lucidrains/DALLE2-pytorch/issues/167
When running example training through rainbow_dalle notebook I managed to generate a dataset and train VAE, but I received error when attempting to train DALL-e:
My environment is as follows: OS: ubuntu 18.04 Python 3.9.13 Torch 1.12.0 torchvision 0.13.0
Have you observed similar issue? What solution could fix it?