Add embedding table compression script #261
Merged
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
Add a script to compress TFRA dynamic embedding tables' value datatype from float32 to float16. It can be used for exported SavedModel.
rhdong
reviewed
Jul 6, 2022
demo/dynamic_embedding/movielens-100k-estimator/movielens-100k-estimator.py
Outdated
Show resolved
Hide resolved
|
Hi @xiangzez , thank you for your first contribution, it's very valuable! I would like to know how much does compression affect the model's accuracy? I mean if you have done some comparison or benchmark, you can add it to the Readme of the demo for other user. Thank you! |
|
Hi @rhdong, this is a work we did for our customers, so we don't have accuracy change data of their production models. However this demo shows comparison of loss before/after compression. I have added example outputs in readme. |
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Outdated suggestions cannot be applied.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.
Description
Brief Description of the PR:
Hi, This PR contains a script to compress TFRA dynamic embedding tables' value datatype from float32 to float16. It can be used to process exported SavedModel and can reduce ~50% size of models that have very large embedding tables. It is integrated into
movielens-100k-estimatordemo and I hope it can be a reference to other users who have similar needs.The script only edits CuckooHashTable related ops in the model graph and will add a cast after CuckooHashTableFind op to keep outputs the same, so other parts of the model are not affected.
Type of change
Checklist:
How Has This Been Tested?
If you're adding a bugfix or new feature please describe the tests that you ran to verify your changes:
*