New metrics for weight compression with dynamic quantization #2829
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
AlexanderDokuchaev
requested changes
Jul 22, 2024
alexsu52
reviewed
Jul 23, 2024
ljaljushkin
force-pushed
the
nl/dyn_quant_conform
branch
2 times, most recently
from
65db012 to
eeb21a0
Compare
|
@alexsu52, please review, but there are still not stable results for GTPQ - 148819 |
ljaljushkin
force-pushed
the
nl/dyn_quant_conform
branch
from
eeb21a0 to
c7f3566
Compare
|
Closing this PR, since dynamic quantization of activations may lead to more unstable accuracy results and there's no |
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Outdated suggestions cannot be applied.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
Use default runtime options for weight compression in conformance tests.
Dynamic Quantization of Activations is enabled by default at least since 2024.4: openvinotoolkit/openvino#25054
KV Cache compression to u8 since 2025: openvinotoolkit/openvino#27454
Reason for changes
Dynamic Quantization Activations and KV Cache compression in U8 should be default options for users since OpenVINO 2025.
Changed tests to validate weight compression with the option.
Related tickets
144846, 146143, 145701
Tests
51 build with openvino-nightly (2025.0.0-17353-acccb227fba)
252 build with current openvino (2024.4.0-16579-c3152d32c9c-releases/2024/4)
