Skip to content
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.

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

Maxpool2D block sharding bug #14249

Closed
mywoodstock opened this issue Oct 24, 2024 · 3 comments
Closed

Maxpool2D block sharding bug #14249

mywoodstock opened this issue Oct 24, 2024 · 3 comments

Comments

@mywoodstock
Copy link
Contributor

Branch: asarje/maxpool-block-bug
Unit test: tests/ttnn/unit_tests/operations/test_maxpool2d.py::test_run_max_pool_block_shard
(Shape = [1, 512, 10, 10], kernel_size = (5,5), padding = (2, 2), stride = (1, 1))

The output PCC is bad -- output tensor seems to have some junk values mixed in.

@punithsekar
Copy link
Contributor

@mywoodstock @wransom-TT, Not only the above mentioned configurations are making issues.
Even the following configurations also give a bad PCC,
(Shape = [1, 512, 10, 10], kernel_size = (9,9), padding = (4, 4), stride = (1, 1))
(Shape = [1, 512, 10, 10], kernel_size = (13,13), padding = (6, 6), stride = (1, 1))

@mywoodstock
Copy link
Contributor Author

@punithsekar yes, they are all the same issue -- basically related to kernel sizes larger then 4x4. We are looking into this.

@punithsekar
Copy link
Contributor

Thanks for the acknowledgment, Abhinav. Since you mentioned specific configurations, I’ve added them.

wransom-TT added a commit that referenced this issue Oct 31, 2024
@mywoodstock mywoodstock added P1 and removed P0 labels Nov 5, 2024
wransom-TT added a commit that referenced this issue Nov 8, 2024
…ns (#14531)

* #14249: Fixed bug for width and block sharding with large kernel sizes and wide reductions
ct-clmsn pushed a commit to ct-clmsn/tt-metal that referenced this issue Nov 12, 2024
…ns (tenstorrent#14531)

* tenstorrent#14249: Fixed bug for width and block sharding with large kernel sizes and wide reductions
Sign up for free to join this conversation on GitHub. Already have an account? Sign in to comment
Projects
None yet
Development

No branches or pull requests

3 participants