9080 expose mat mul precision #9081
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
bhavya01
reviewed
May 2, 2025
ysiraichi
reviewed
May 5, 2025
tengyifei
approved these changes
May 5, 2025
bhavya01
approved these changes
May 6, 2025
vlasenkoalexey
approved these changes
May 9, 2025
zhanyong-wan
pushed a commit
to zhanyong-wan/pytorch-xla
that referenced
this pull request
May 9, 2025
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Outdated suggestions cannot be applied.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 added a binding to get the mat mul precision to the init_python_bindings.cpp
I exposed it as a brand new module, torch_xla.backends. This is an ackward name but the goal is to eventually migrate this to torch.backends.xla, to parallel torch.backends.{cuda, cpu, mps, etc}.
I got advice from on very exact numerics for default (1 pass). I made estimates on the 3 pass and 6 pass technique.
I was careful to ensure that there was a non-zero delta from the torch64 cpu calculation - it's easy to make a mistake and end up with your "reference" math also be rounded.
I have a detailed guide on this coming after this PR goes in.