Making task 1 run on a laptop.

Written on 03.03.2024 15:24 by Michael Schott

Hi all,

since it has occurred multiple times now that teams ran notebooks on the submit node, we want to make sure this is not caused by the computational load required to solve task 1 exceeding your computational resources. Note that task 1 can be run without problems on a laptop with 8 GB of ram and possibly less. However, for this you have to disable the gradient computation, since otherwise PyTorch stores the gradients for all weights in each forward pass, filling up your memory very quickly. To prevent PyTorch from doing this, you can e.g. wrap the code for the forward pass in a torch.no_grad() context, such as

with torch.no_grad():

    # your code here

Should this not work for you, or you face other issues in running the notebook for task 1, please let us know in the forum.


NNTI Tutors.

Privacy Policy | Legal Notice
If you encounter technical problems, please contact the administrators.