Use latest optimum to not call deprecated shared_memory API

[ljaljushkin]

Changes

use the latest release of optimum-intel==1.15.2

Reason for changes

The previously used version of optimum had a call of deprecated API that was removed in 2024 release of OpenVINO. https://github.com/openvinotoolkit/openvino/issues/21979

outputs = self.request(inputs, shared_memory=True)

Latest optimum migrated to the actual API.

Related tickets

n/a

Tests

• [x] weight compression conformance (local run)
• [x] weight compression conformance (CI build 23)
• [ ] ptq conformance build (CI build 320)

[codecov[bot]]

Codecov Report

All modified and coverable lines are covered by tests :white_check_mark:

Project coverage is 84.83%. Comparing base (12f4720) to head (52dbd42).

Additional details and impacted files

@@             Coverage Diff             @@
##           develop    #2569      +/-   ##
===========================================
- Coverage    91.18%   84.83%   -6.36%     
===========================================
  Files          492      492              
  Lines        45096    45100       +4     
===========================================
- Hits         41123    38261    -2862     
- Misses        3973     6839    +2866     

see 56 files with indirect coverage changes

Flag	Coverage Δ
COMMON	43.92% <ø> (-0.01%)	:arrow_down:
ONNX	34.66% <ø> (-0.01%)	:arrow_down:
OPENVINO	∅ <ø> (∅)
TENSORFLOW	29.87% <ø> (-0.01%)	:arrow_down:
TORCH	65.76% <ø> (-0.01%)	:arrow_down:

Flags with carried forward coverage won't be shown. Click here to find out more.

Components	Coverage Δ
common	93.11% <ø> (-0.66%)	:arrow_down:
torch	93.59% <ø> (ø)
tensorflow	93.74% <ø> (ø)
onnx	93.05% <ø> (ø)
openvino	25.58% <ø> (-68.49%)	:arrow_down:
ptq	69.78% <ø> (-20.45%)	:arrow_down:

[ljaljushkin]

run tensorflow pre-commit tests

[ljaljushkin]

PTQ test crashes with segfault, reproduced locally. Please don't merge until resolving the issue @KodiaqQ

[ljaljushkin]

gfault, reproduced locally. Please don't merge until resolving

Created a bug 135375 for CPU plugin

[KodiaqQ]

@ljaljushkin, the bug is still actual and needs to be fixed. What will we do with the segfault?

[ljaljushkin]

@ljaljushkin, the bug is still actual and needs to be fixed. What will we do with the segfault?

This is the bug for CPU plugin, therefore we will wait for the fix.

[KodiaqQ]

This is the bug for CPU plugin, therefore we will wait for the fix.

We will wait till the next OpenVINO release (or even another one) until then the weekly runs will be red. Am I right?

[ljaljushkin]

e the latest release of optimum-intel==1.15.

No, originally you asked not correct question. Weekly tests are failing not because of segfault, they fail because test depend on the old optimum which uses a removed API. Segfault coming from using latest openvino and optimum version. I will think about another approach for fixing weekly test to avoid bug with segfault, while runtime team is working on fixing it.

[KodiaqQ]

No, originally you asked not correct question. Weekly tests are failing not because of segfault, they fail because test depend on the old optimum which uses a removed API. Segfault coming from using latest openvino and optimum version. I will think about another approach for fixing weekly test to avoid bug with segfault, while runtime team is working on fixing it.

I push this because the weekly tests are red for a few weeks and there is no estimation to merge this PR due to segfault. And there is even no activity with the segfault itself. Red weekly tests affect the upcoming NNCF release as well. It's highly recommended to fix it before the release branch creation.

[alexsu52]

No, originally you asked not correct question. Weekly tests are failing not because of segfault, they fail because test depend on the old optimum which uses a removed API. Segfault coming from using latest openvino and optimum version. I will think about another approach for fixing weekly test to avoid bug with segfault, while runtime team is working on fixing it.

I push this because the weekly tests are red for a few weeks and there is no estimation to merge this PR due to segfault. And there is even no activity with the segfault itself. Red weekly tests affect the upcoming NNCF release as well. It's highly recommended to fix it before the release branch creation.

As workaround you can use stateless model instead of stateful.

[ljaljushkin]

Here's the fix for segfault. https://github.com/openvinotoolkit/openvino/pull/23581