Skip to content

Fixes in helion puzzles #1104

New issue

Have a question about this project? Sign up for a free GitHub account to open an issue and contact its maintainers and the community.

Sign up for GitHub

By clicking “Sign up for GitHub”, you agree to our terms of service and privacy statement. We’ll occasionally send you account related emails.

Already on GitHub? Sign in to your account

Jump to bottom
Merged
merged 10 commits into from
Nov 9, 2025
Merged

Fixes in helion puzzles #1104

merged 10 commits into from
Nov 9, 2025

Conversation

@Athe-kunal
Copy link
Contributor

@Athe-kunal Athe-kunal commented Nov 8, 2025

  1. Fixes in the softmax operation (max is not supported)
  2. For the convolution kernel, I fixed the shape and devices issue, but the result does not match with conv2d_spec

@Athe-kunal
Enable gradient tracking for 'y' in mul_relu_block_back_spec
4039022 
I was getting an error. Resolved after enabling gradient checking for y

```
RuntimeError: One of the differentiated Tensors does not require grad
```
@Athe-kunal
Merge pull request #1 from Athe-kunal/Athe-kunal-patch-1
892ca0c 
Enable gradient tracking for 'y' in mul_relu_block_back_spec
@Athe-kunal
Merge branch 'pytorch:main' into main
14422cc
@Athe-kunal
Update accumulator initialization in helion_puzzles.rst
59a621f 
Change accumulator initialization to use the device of input tensor. And use `zeroes` instead of `zeroes_like` because the alter accepts tensor
@Athe-kunal
Merge branch 'pytorch:main' into main
96a887e
@Athe-kunal
Replace torch.max with torch.amax
50d4e51 
torch.max is not supported for helion kernels
@Athe-kunal
Merge branch 'pytorch:main' into main
9e598c0
@Athe-kunal
Modify tensor initialization and padding for device compatibility
cd0fb69 
Updated tensor operations to ensure compatibility with input device.
@meta-cla meta-cla bot added the CLA Signed This label is managed by the Meta Open Source bot. label Nov 8, 2025
jansel
jansel requested changes Nov 8, 2025
View reviewed changes
Copy link
Contributor

@jansel jansel left a comment

Choose a reason for hiding this comment

The reason will be displayed to describe this comment to others. Learn more.

That for the fix, just a minor nit.

@Athe-kunal
Copy link
Contributor Author

Athe-kunal commented Nov 8, 2025 •
edited
Loading

@jansel
The testcases for the Conv2D puzzle is failing though. I am trying to debug, but I am not sure which part is going wrong here

AssertionError: Tensor-likes are not close!

Mismatched elements: 256 / 256 (100.0%)
Greatest absolute difference: 13.31396770477295 at index (0, 2, 3) (up to 1e-05 allowed)
Greatest relative difference: inf at index (1, 0, 7) (up to 1.3e-06 allowed)

@jansel jansel merged commit 5d0cd02 into pytorch:main Nov 9, 2025
14 of 15 checks passed
@jansel
Copy link
Contributor

jansel commented Nov 9, 2025

It looks like the test passed in CI

@Athe-kunal
Copy link
Contributor Author

Athe-kunal commented Nov 9, 2025

Hmm I see, probably some issues with the dtypes

Sign up for free to join this conversation on GitHub. Already have an account? Sign in to comment

Reviewers

@jansel jansel jansel approved these changes

Assignees

No one assigned

Labels

CLA Signed This label is managed by the Meta Open Source bot.

Projects

None yet

Milestone

No milestone

Development

Successfully merging this pull request may close these issues.

2 participants

@Athe-kunal @jansel