Use regular division inside Scale Estimation #3210
Open
+6
−27
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
github-actions
bot
added
NNCF OpenVINO
Comment on lines
38
to
42
| tinyllama_scale_estimation_per_channel_backend_TORCH: | ||
| metric_value: 0.81389 | ||
| metric_value: 0.80873 | ||
| num_int4: 188 | ||
| num_int8: 124 | ||
| atol: 0.006 # difference across devices: 0.80873 vs 0.81389 |
There was a problem hiding this comment.
A reference value for PT is changed to stay aligned with OV. Interestingly, the new value for OV (0.80873) is also the value that PT backend gets on some devices.
nikita-savelyevv
requested review from
andreyanufr,
alexsu52 and
AlexanderDokuchaev
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Outdated suggestions cannot be applied.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.
Changes
Compute division inside SE algorithm always as
a/binstead ofa*(1/b)in some cases.Reason for changes
During implementation #2727 some choices were made regarding how division operation is computed in order for the changes to be completely aligned with the previous implementation. Namely, before #2727 some divisions were computed as
a*(1/b), and this is currently still the case.The way these divisions are computed originally was not intended. Now, all divisions are aligned to the
a/bform.Related tickets
163286
Tests