disable kv cache compression for fp vlm #1080
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
eaidova
force-pushed
the
ea/enable_rt_info_for_vlm
branch
from
ede24e1 to
33cef0f
Compare
|
The docs for this PR live here. All of your documentation changes will be reflected on that endpoint. The docs are available until 30 days after the last update. |
AlexKoff88
approved these changes
Dec 19, 2024
nikita-savelyevv
approved these changes
Dec 19, 2024
|
@echarlaix @IlyasMoutawwakil could you please take a look? ov 2024/6 release happened couple of hours ago, so this minicpmv test failure should be visible on main branch |
3 tasks
AlexKoff88
pushed a commit
that referenced
this pull request
Dec 23, 2024
* Support AWQ models * Add tests * Add dependencies * Fix tests * enable awq export only if ov support it * fix style (#2) * disable awq and gptq install for old torch (#3) * fix style * disable autogptq and autoawq install for old transformers testing * separate common quant models patching and gptq (#4) * disable windows install (#5) * separate common quant models patching and gptq * disable awq windows * skip logits check for quantized models (#6) * fix test after rebase * fix testing condition for 2024.6 and unpatch in case if failed * Fix qwen2-vl tests (#1084) * Skip private mdoel loading test for external contributors (#1082) * Fix reshaping unet if timestep is 0d tensor (#1083) * Disable kv cache compression for fp vlm (#1080) * Support AWQ models * Add tests * Add dependencies * Fix tests * enable awq export only if ov support it * fix style (#2) * disable awq and gptq install for old torch (#3) * fix style * disable autogptq and autoawq install for old transformers testing * separate common quant models patching and gptq (#4) * disable windows install (#5) * separate common quant models patching and gptq * disable awq windows * skip logits check for quantized models (#6) * fix test after rebase * fix testing condition for 2024.6 and unpatch in case if failed * add necessary packages in test_openvino_full * fix code style after rebase (#7) --------- Co-authored-by: eaidova <[email protected]> Co-authored-by: Nikita Savelyev <[email protected]> Co-authored-by: Ella Charlaix <[email protected]>
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.
What does this PR do?
Fixes issue with failed minicpmv with ov nightly. starting from 2024.6 openvino will use kv cache compression by default enabled, that may impact model accuracy, but identify when it should be disabled can not be predicted on runtime level, so we proposed addition of specific hint for such models (by our agreement it should be done for noncompressed models only) - extended this approach to handle language models as part visual language models
Before submitting