Thanks to visit codestin.com
Credit goes to github.com

Skip to content

Add minimal [segment|grouped]_matmul CPU implementation #111

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 4 commits into from
Oct 3, 2022
Merged

Add minimal [segment|grouped]_matmul CPU implementation #111

merged 4 commits into from
Oct 3, 2022

Conversation

@DamianSzwichtenberg
Copy link
Member

@DamianSzwichtenberg DamianSzwichtenberg commented Sep 23, 2022 •
edited
Loading

Added minimal [segment|grouped]_matmul CPU implementation, using at::matmul_out.

Next steps (other PR):
Optimize code using gemm_batched (preview).

Update:
An additional fix was provided for grouped_matmul which occurred on both CPU and GPU.

C++ exception with description "There were no tensor arguments to this function (e.g., you passed an empty list of Tensors), but no fallback function is registered for schema pyg::grouped_matmul.  This usually means that this function requires a non-empty list of Tensors, or that you (the operator writer) forgot to register a fallback function.

It was caused by the wrong grouped_matmul function signature - it was not aligned with the schema provided at TORCH_LIBRARY_FRAGMENT call. As stated here, Tensor[] as input is equivalent of at::TensorList.

@DamianSzwichtenberg DamianSzwichtenberg self-assigned this Sep 23, 2022
@codecov-commenter
Copy link

codecov-commenter commented Sep 25, 2022 •
edited
Loading

Codecov Report

Merging #111 (7cadd0a) into master (285bd21) will increase coverage by 5.10%.
The diff coverage is 100.00%.

❗ Current head 7cadd0a differs from pull request most recent head a642ab8. Consider uploading reports for the commit a642ab8 to get more accurate results

@@            Coverage Diff             @@
##           master     #111      +/-   ##
==========================================
+ Coverage   89.23%   94.33%   +5.10%     
==========================================
  Files          16       19       +3     
  Lines         418      477      +59     
==========================================
+ Hits          373      450      +77     
+ Misses         45       27      -18
Impacted Files Coverage Δ
pyg_lib/csrc/ops/matmul.h 100.00% <ø> (ø)
pyg_lib/csrc/ops/cpu/matmul_kernel.cpp 100.00% <100.00%> (ø)
pyg_lib/csrc/ops/matmul.cpp 91.83% <100.00%> (+73.65%) ⬆️
pyg_lib/csrc/utils/check.cpp 100.00% <100.00%> (ø)
pyg_lib/csrc/sampler/neighbor.cpp 100.00% <0.00%> (ø)
pyg_lib/csrc/sampler/cpu/neighbor_kernel.cpp 89.71% <0.00%> (+0.33%) ⬆️
pyg_lib/csrc/utils/convert.cpp 100.00% <0.00%> (+100.00%) ⬆️

📣 We’re building smart automated test selection to slash your CI/CD build times. Learn more

@DamianSzwichtenberg
Copy link
Member Author

@rusty1s It seems that docs build is failing because of:

File "/home/docs/checkouts/readthedocs.org/user_builds/pyg-lib/envs/111/lib/python3.8/site-packages/sphinx_rtd_theme/layout.html", line 13, in top-level template code
    {%- set (_ver_major, _ver_minor, _ver_bugfix) = sphinx_version.split('.') | map('int') -%}
ValueError: too many values to unpack (expected 3)

The installed version is now 5.2.0.post0... Maybe we should specify the version explicitly in requirements.txt?

@rusty1s
Copy link
Member

rusty1s commented Sep 25, 2022

Yes, will look into.

@rusty1s
Copy link
Member

rusty1s commented Sep 26, 2022

Seems like sphinx==5.2.1 fixed this issue.

@DamianSzwichtenberg DamianSzwichtenberg force-pushed the minimal-mm-cpu-impl branch 2 times, most recently from a1f7eda to 7cadd0a Compare September 26, 2022 08:31
@DamianSzwichtenberg
Copy link
Member Author

@rusty1s Please take a look. 😉

rusty1s
rusty1s approved these changes Oct 3, 2022
View reviewed changes
Copy link
Member

@rusty1s rusty1s left a comment

Choose a reason for hiding this comment

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

Thanks for this PR and the dispatcher fix. I really had no idea how one could resolve it :)

@rusty1s rusty1s enabled auto-merge (squash) October 3, 2022 16:57
@rusty1s rusty1s merged commit 9832ea6 into pyg-team:master Oct 3, 2022
@puririshi98 puririshi98 mentioned this pull request Oct 3, 2022
Merged
Sign up for free to join this conversation on GitHub. Already have an account? Sign in to comment

Reviewers

@rusty1s rusty1s rusty1s approved these changes

Assignees

@DamianSzwichtenberg DamianSzwichtenberg

Labels

0 - Priority P0 feature ops

Projects

None yet

Milestone

No milestone

Development

Successfully merging this pull request may close these issues.

3 participants

@DamianSzwichtenberg @codecov-commenter @rusty1s