Allow for matrix-matrix and matrix-vector products with KFACLinearOperator and KFACInverseLinearOperator without converting to numpy
#91
Conversation
This file contains bidirectional Unicode text that may be interpreted or compiled differently than what appears below. To review, open the file in an editor that reveals hidden Unicode characters.
Learn more about bidirectional Unicode characters
![sourcery-ai[bot]](https://avatars.githubusercontent.com/in/48477?s=60&v=4){kind=small}
There was a problem hiding this comment.
Hey @runame - I've reviewed your changes and they look great!
General suggestions:
- Consider refactoring input validation and device compatibility checks into separate methods to improve code readability and maintainability.
- Explore more efficient approaches for tensor concatenation to optimize performance.
- Enhance error messages with more specific details to aid in debugging and improve the developer experience.
Here's what I looked at during the review
- 🟡 General issues: 4 issues found
- 🟢 Security: all looks good
- 🟢 Testing: all looks good
- 🟢 Complexity: all looks good
- 🟢 Docstrings: all looks good
Thanks for using Sourcery. We offer it for free for open source projects and would be very grateful if you could help us grow. If you like it, would you consider sharing Sourcery on your favourite social media? ✨
curvlinops/kfac.py
Outdated
Comment on lines
346
to
347
| if return_tensor: | ||
| M_torch = cat([rearrange(M, "k ... -> (...) k") for M in M_torch], dim=0) |
There was a problem hiding this comment.
suggestion (performance): Consider the efficiency of tensor concatenation.
Concatenating tensors in a loop can be inefficient, especially for large numbers of tensors. It might be beneficial to explore alternative approaches that could reduce the computational overhead, such as preallocating a tensor of the correct size and filling it.
Co-authored-by: sourcery-ai[bot] <58596630+sourcery-ai[bot]@users.noreply.github.com>
Pull Request Test Coverage Report for Build 8360156180Details
💛 - Coveralls |
runame
changed the title
KFACLinearOperator without converting to numpyKFACLinearOperator and KFACInverseLinearOperator without converting to numpy
f-dangel
reviewed
Mar 18, 2024
f-dangel
reviewed
Mar 19, 2024
f-dangel
reviewed
Mar 20, 2024
Sign up for free
to join this conversation on GitHub.
Already have an account?
Sign in to comment
Add this suggestion to a batch that can be applied as a single commit.
This suggestion is invalid because no changes were made to the code.
Suggestions cannot be applied while the pull request is closed.
Suggestions cannot be applied while viewing a subset of changes.
Only one suggestion per line can be applied in a batch.
Add this suggestion to a batch that can be applied as a single commit.
Applying suggestions on deleted lines is not supported.
You must change the existing code in this line in order to create a valid suggestion.
Outdated suggestions cannot be applied.
This suggestion has been applied or marked resolved.
Suggestions cannot be applied from pending reviews.
Suggestions cannot be applied on multi-line comments.
Suggestions cannot be applied while the pull request is queued to merge.
Suggestion cannot be applied right now. Please check back later.
Addresses #71, but only for the
KFACLinearOperatorandKFACInverseLinearOperator.KFACLinearOperator/KFACInverseLinearOperatoris arguably the only linear operator here that is likely to be used for preconditioned-gradient methods for large scale neural networks. Therefore, addingtorch_matmatandtorch_matvecmethods seem like a simple solution to avoid unnecessary device transfers, which are a bottleneck for this use case. However, this doesn't address the issue in general.