Quantization Fixes #35
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
dbogunowicz
approved these changes
Apr 25, 2024
| :param observed: observed tensor to calculate quantization parameters for | ||
| :return: tuple of scale and zero point derived from the observed tensor | ||
| """ | ||
|
|
||
| min_val = torch.tensor([observed.min()]) | ||
| max_val = torch.tensor([observed.max()]) | ||
| min_val, max_val = torch.aminmax(observed) |
There was a problem hiding this comment.
I wonder why this makes a difference
bfineran
approved these changes
Apr 25, 2024
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Outdated suggestions cannot be applied.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.
A few change so we match pytorch quantization more closely:
Before these changes we were consistently seeing a 10-15% drop in perplexity compared to pytorch quantization. Now we are matching them within 2% and aren't consistently worse. We still don't have exact matching of scale/zeropoint calculations but I'm chalking that up to C++ vs Python differences (some of the pytorch quantization is implemented in c++)
perplexity test is in: neuralmagic/sparseml#2246