Set cache_precision = weights_precision in TBE if it is not explicitly set #3370
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
|
This pull request was exported from Phabricator. Differential Revision: D65865527 |
✅ Deploy Preview for pytorch-fbgemm-docs ready!
To edit notification comments on pull requests, go to your Netlify site configuration. |
…y set (pytorch#3370) Summary: X-link: facebookresearch/FBGEMM#461 For historical reasons, we have decoupled cache_precision and weights_precision. This was to allow for lower precision embedding, while the hot ones in the cache are using higher precision. But it is not used. This decoupling has been source of unintentional difference in cache_precision and weights_precision for typical EMO usecases where we do want cache_precision == weights_precision. We had enforced this in torchrec (see [this](https://www.internalfb.com/code/fbsource/[3868325cdafd]/fbcode/torchrec/distributed/batched_embedding_kernel.py?lines=962-963%2C1446-1447)), but as new stacks are enabled, this could be overwritten. Here we enforce cache_precision = weights_precision if cache precision is not explicitly set. Differential Revision: D65865527
ehsanardestani
force-pushed
the
export-D65865527
branch
from
1442708
to
dceb83d
Compare
|
This pull request was exported from Phabricator. Differential Revision: D65865527 |
|
This pull request has been merged in 10ae4f8. |
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Outdated suggestions cannot be applied.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.
Summary:
X-link: https://github.com/facebookresearch/FBGEMM/pull/461
For historical reasons, we have decoupled cache_precision and weights_precision. This was to allow for lower precision embedding, while the hot ones in the cache are using higher precision. But it is not used.
This decoupling has been source of unintentional difference in cache_precision and weights_precision for typical EMO usecases where we do want cache_precision == weights_precision. We had enforced this in torchrec (see this), but as new stacks are enabled, this could be overwritten.
Here we enforce cache_precision = weights_precision if cache precision is not explicitly set.
Differential Revision: D65865527