[QwixQuantizedTraining] Clip Gradients for Out-of-Bound Activations and Weights in dot_general_qt.#173
Merged
Merged
Conversation
354ea7b to
91f52ce
Compare
c759bf4 to
2ae0179
Compare
…nd Weights in dot_general_qt. This change zeroes out the backward pass gradients (dlhs, drhs) that correspond to forward pass inputs (both activations and weights) that were outside their calibration bounds. - For "in-bound" values, qwix QT will calculate the derivative using low precision data type. - For "out-bound" values, qwix QT will set the derivative to 0. This will ensure mathematical correctness and numerical stability. PiperOrigin-RevId: 833558845
2ae0179 to
eb06932
Compare
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
1 participant
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Outdated suggestions cannot be applied.This suggestion has been applied or marked resolved.Suggestions cannot be applied from pending reviews.Suggestions cannot be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.
[QwixQuantizedTraining] Clip Gradients for Out-of-Bound Activations and Weights in dot_general_qt.
This change zeroes out the backward pass gradients (dlhs, drhs) that correspond to forward pass inputs (both activations and weights) that were outside their calibration bounds.
This will ensure mathematical correctness and numerical stability.