search

namiyousef / multi-task-learning

Repository for multi task learning
2 stars 2 forks source link

Type bug when trying GPU #10

Closed namiyousef closed 2 years ago

namiyousef commented 2 years ago

There's a bug due to some ype mismatch, torch.FloatTensor vs. torch.cuda.FloatTensor. There is a notebook file that you should be able to run on colab.

More details, see here

namiyousef commented 2 years ago

Found bug. The bug is due to _input_conv method for the ResU body. Because the layers are instantiated outside of an nn.Module class, they aren't bound to it. Therefore, sending the class to GPU does not suffice. Fixing this will require a re-writing of the code, I will be working on that. In the meantime, I believe @ConnorWatts's new branch (e.g. final) with the resnet encoder should work, since the issue seems to be stemming from the encoder. There may be further issues as I haven't had time to set all the other stuff (e.g. issues with testing, or issues with calculation of loss functions). If any of you have time, please try running the new notebook (Please see issue #4 for how to run it). If you get any bugs, please take images of the entire logs and add them as comments to this thread.

namiyousef commented 2 years ago

Change milestone to prepare code for submission