RuntimeError: CUDA error: initialization error

Deep Learning with PyTorch: Zero to GANs Course Project
,
#1

Hello

I have done the project yesterday using a RestNet Model and it was working very good. But today I was writting some describtions on it, and then it does not work any more!!

I get this error by show_denormalized_batch() function. And the same Error is comming later also when I make a training round too!

Anyone knows what is happening and how to solve this issue?

Thanks in advance

Greatings
Alaa

@aakashns
@aakashshankar

image
image1706×1088 169 KB

image
image1686×1010 151 KB

#2

Hey @alaa-diab7777 may I know in which environment are you working on? Also, can you share the whole notebook? There are similar issues on github either pointing to num_workers or transferring of dataloader to GPU

#3

Hey,
I was working in Colab. This is correct, the problem appears also when loading the data into GPU too.

I do changed the structure of my network and dimensions of pooling, then the error is surprisingly gone. :slight_smile: I still do not know how it happened.

Here is the link for my notebook

Feel free to check it, and thanks for the reply