Remove the clamp op when we do symmetric quantization on a tensor#9465
Merged
vanbasten23 merged 3 commits intomasterfrom Jul 10, 2025
Merged
Remove the clamp op when we do symmetric quantization on a tensor#9465vanbasten23 merged 3 commits intomasterfrom
vanbasten23 merged 3 commits intomasterfrom
Conversation
373f791 to
69fd267
Compare
vanbasten23
commented
Jul 9, 2025
Collaborator
|
Looks like the tolerance of the numerical error is relaxed in a few tests, so this implies the new impl is generating bigger error? |
Collaborator
Author
Good question. I explained here #9465 (comment). I first run the tests without any changes on v5e and the test |
lsy323
approved these changes
Jul 9, 2025
69fd267 to
268b19b
Compare
Collaborator
Author
|
Thanks for the review! |
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
Sign up for free
to join this conversation on GitHub.
Already have an account?
Sign in to comment
2 participants
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Outdated suggestions cannot be applied.This suggestion has been applied or marked resolved.Suggestions cannot be applied from pending reviews.Suggestions cannot be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.
Originally, when we do symmetric quantization, we do
x/scale = x / (max(abs(x)) / 127) = (x / max(abs(x)) * 127
Mathematically, we know that
x / max(abs(x)always returns the range of -1 to 1, meaningx / scalealways returns the range of -127 to 127. Therefore, torch.clamp can be skipped.Idea credit to Kyuyeun Kim @kyuyeunk
Test plan: