Issues for Beginners/Students

kaad01 · June 23, 2022, 1:24pm

Hi, I am a Computer Science Student, and my study partner and I would like to contribute to PyTorch. What would be a good issue for beginners?
Our machine learning knowledge is very solid, but our goal is to submit a pull request which gets accepted in the next two weeks or so. So an easier Issue would be more fitting for now.

Thank you!

ptrblck · June 24, 2022, 5:07am

Hi Kaan!

I think you and your study partner could take a look at the “good first issue” section in GitHub and maybe you can find something interesting there.
For a general guide to contribute to PyTorch, check:

the Contribution Guide
these Contributing docs
and this video/blog showing how to “fix your first PyTorch bug” by @tom

EDIT: it might also be interesting to take a look at this “tracing with primitives” update from ~1h ago, which is also open for contributions, so let me know if anything would fit your interest.

kaad01 · June 25, 2022, 2:26am

Thank you very much, we will look into it.

kaad01 · July 2, 2022, 12:01am

Hey,
I’m ready to start now.
What do you think of this Issue:

github.com/pytorch/pytorch

`xlogy` returns wrong gradient when `input` is 0

opened 01:12PM - 01 Jul 22 UTC

upupyy

### 🐛 Describe the bug `xlogy` returns wrong gradient when `input` is 0. Since …it is continuous and differential at the point 0, it shouldn't return 0 as the gradient. ```py import torch def fn(input): fn_res = input.xlogy(2) return fn_res input = torch.tensor([[0., 0., 0., 0.]], dtype=torch.float64, requires_grad=True) torch.autograd.gradcheck(fn, (input)) ``` ``` torch.autograd.gradcheck.GradcheckError: Jacobian mismatch for output 0 with respect to input 0, numerical:tensor([[0.6931, 0.0000, 0.0000, 0.0000], [0.0000, 0.6931, 0.0000, 0.0000], [0.0000, 0.0000, 0.6931, 0.0000], [0.0000, 0.0000, 0.0000, 0.6931]], dtype=torch.float64) analytical:tensor([[0., 0., 0., 0.], [0., 0., 0., 0.], [0., 0., 0., 0.], [0., 0., 0., 0.]], dtype=torch.float64) ``` ### Versions pytorch: 1.11.0

And how do you find the right File to look at? For example, in this issue, I have to look into the xlogy function. But where do find it?

Thank you for your help

ptrblck · July 2, 2022, 12:26am

Yes, it might be a good task. I’ve added the links to the GitHub issue.

oke-aditya · July 2, 2022, 6:36pm

Hi! Not sure if this issue is specific to PyTorch
Torchvision too has beginner friendly issues Issues · pytorch/vision · GitHub

Or Some issues for bit intermediate users like [RFC] Batteries Included - Phase 2 · Issue #5410 · pytorch/vision · GitHub