Add scaling fusion analysis and update docs #21
Merged
Conversation
This file contains hidden or 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
DouglasOrr
approved these changes
May 23, 2023
|
Thanks v much! Changes are here - nothing major, mostly just removing all the caveats about the bugged bwd pass |
DouglasOrr
approved these changes
May 23, 2023
| For a thorough analysis of the effect of unit-scaling | ||
| (for both individual operations and larger blocks), see the | ||
| `benchmarking compiled unit-scaled ops <https://github.com/graphcore-research/unit-scaling/tree/main/analysis/benchmarking_compiled_unit_scaled_ops.ipynb>`_ notebook. | ||
| Note that there's a bug in the latest PyTorch version meaning the backward pass |
There was a problem hiding this comment.
Might be good to give a version number here too (is it 2.0.1)?
Closed
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Outdated suggestions cannot be applied.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.
I've added the torch.compile analysis notebook to the repo and updated our docs to explain the recommended approach to optimisation/fusion. No need to go too deep on the notebook, but feel free to look at my conclusions there!
Also fyi, my issue flagging this up to the PyTorch people: pytorch/pytorch#101937