Changes from upstream/master #2
Open
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
category: build
category: C API
category: CI
github_actions
category: Core
category: CPP API
category: CPU
category: dependency_changes
category: docs
category: GPU
category: NPU
category: IE Tests
category: inference
category: packaging
category: Python API
category: JS API
category: TEMPLATE
category: TF FE
category: TFL FE
category: PyTorch FE
no-match-files
labels
…8085) ### Details: - Add imports from _pyopenivno on the top of the openvino `__init__.py` - Move `Node` binary operators support at the bottom of openvino `__init__.py` Classes and functions available in `openvino/runtime` will match those in flatten `openvino` ### Tickets: - [CVS-129453](https://jira.devtools.intel.com/browse/CVS-129453) --------- Signed-off-by: Alicja Miloszewska <[email protected]> Co-authored-by: Michal Lukaszewski <[email protected]>
### Details: - *Fail if non continuous tensor is set* ### Tickets: --------- Signed-off-by: Bogdan Pereanu <[email protected]>
…#28091) Currently 'scale' is obtained using a ShapeOf expression as the hidden_dim may be dynamic in some cases and not propagated, so we can't use it directly to create a 'scale' Constant. Check if hidden_dim is static and use it to calculate 'scale' directly omitting the ShapeOf expression. Ticket: * [CVS-158394](https://jira.devtools.intel.com/browse/CVS-158394) Signed-off-by: Andrii Staikov <[email protected]>
…8105) ### Details: - The model after read by IR frontend shows models output tensors names as Nodes connected to Result's. But by default this names are not dedicated Results name which can cause during pre-post processing that names stay on node and will disapear as model output names. To fix set the names as Results names so during transformations they will stay as model's output names. - Onnx frontens set OV model output's names when converting model to OV represenation. - Fix NPU test which reports `Attempt to get a name for a Tensor without names` ### Related PRs: - #28102 ### Tickets: - CVS-159401 --------- Signed-off-by: Raasz, Pawel <[email protected]>
### Details: - Added variants of Philox algorithm generators from PyTorch and Tensorflow to align generated output for the same seed - Added x64 kernels for Philox (Tensorflow), Mersenne (PyTorch) - Added tests for exactness of results - Spec with changes description available here: #26142 ### Tickets: - 134606 --------- Signed-off-by: PiotrKrzem <[email protected]> Co-authored-by: Michal Lukaszewski <[email protected]>
### Details: - Added reference impl of STFT op with unit and functional tests. ### Tickets: - *CVS-147162* --------- Co-authored-by: Pawel Raasz <[email protected]> Co-authored-by: Michal Lukaszewski <[email protected]>
…in (#28111) **Details:** Stabilize tests for unary operation defined on full real domain **Ticket:** TBD --------- Signed-off-by: Kazantsev, Roman <[email protected]>
### Details: - Should mitigate things like in https://github.com/openvinotoolkit/openvino/actions/runs/12081540979/job/33691617750 ### Tickets: - *158474* --------- Co-authored-by: Mikhail Ryzhov <[email protected]>
### Details: - This PR enables execution FullyConnected operations via OneDNN Matmul Primitive - Matmul_weights_decompression tests are splitted on x64 and arm instances, ARM tests run well via ref matmul. - Newly added functionality is still under debug caps. To try it out: -- Build OV with: -DENABLE_DEBUG_CAPS=ON cmake option -- export OV_CPU_ENABLE_DNNL_MAMTUL_FOR_FC=1
### Details: - Applied OPENVINO_MATCHER_PASS_RTTI helper macro in Plugins. ### Tickets: - CVS-159567 --------- Signed-off-by: Tomasz Jankowski <[email protected]> Co-authored-by: Michal Lukaszewski <[email protected]>
### Details: - Adds RTTI definition for passes derived from ov::pass::ModelPass. - Applies the macro where applicable. ### Tickets: - CVS-159568 --------- Signed-off-by: Tomasz Jankowski <[email protected]> Co-authored-by: Michal Lukaszewski <[email protected]>
…xecutor (#26967) ### Details: - Improve performance of STFT for CPU plugin by reusage RDFT jit Executor - Use parallel loops in stft - No changes in the logic of the existing RDFT executor, RDFTExecutor::build function has been added to keep the RDFT details hidden in cpp as is. - Perf numbers collected within the ticket ### Tickets: - 156115 --------- Co-authored-by: Michal Lukaszewski <[email protected]>
### Details: - *The PR enabled Snippets tokenization for ops `Ceiling` (activation) and `FloorMod` (eltwise):* - *Updated tokenization callback* - *Registered the operations with the corresponding emitters in `CPUGenerator`* - *Added tests* - *This PR is example for the further GFIs'* ### Tickets: - *N/A*
### Details: - update openvino-node package version to 2024.6.0 - update openvino-node to 2024.6.0 in samples
…ions (#28150) ### Details: - Applied OPENVINO_MATCHER_PASS_RTTI helper macro in transformations/op_convertions. ### Tickets: - CVS-159567 --------- Signed-off-by: Tomasz Jankowski <[email protected]>
### Details: - Follow-up from #28077 for an additional comment
…P is using (#28120) ### Details: - Removed a redundant memory copying for data types less than 4 bytes when MMAP is using - Implementation verified by a previously added tests named onnx_external_data_* ### Tickets: - 159161
### Details:
- Implement `__enter__` and `__exit__` to create class-based context
manager
- Remove inheritance from ModelBase in ie_api.Model and make it an
attribute.
- add private property __model that stores `_pyopenvino.Model`. In
`_pyopenvino` such attribute can be accessed as `_Model__model` because
of name mangling
- update pyAPI methods that have `std::shared_ptr<ov::Model>` in their
signatures.
- add `test_model_with_statement` and `test_model_tempdir_fails`
### Motivation:
On Windows reading `ov.Model` from temporary directory leads to
`PermissionError`:
```python
mem_model = generate_model_with_memory(input_shape=Shape([2, 1]),
data_type=Type.f32)
with tempfile.TemporaryDirectory() as model_save_dir:
save_model(mem_model, f"{model_save_dir}/model.xml")
model = Core().read_model(f"{model_save_dir}/model.xml")
```
### Tickets:
- CVS-106987
---------
Signed-off-by: Alicja Miloszewska <[email protected]>
### Details: - Replace imports of openvino.runtime in openvino module ### Tickets: Part of [openvino.runtime deprecation](https://jira.devtools.intel.com/browse/CVS-129450) Signed-off-by: Alicja Miloszewska <[email protected]>
### Details: - *item1* - *...* ### Tickets: - *ticket-id*
…n to RandomUniform-8 (#26142) ### Details: - Added new alignment mode description - Added explanation of Mersenne-Twister algorithm with examples ### Tickets: - None
- Add a command line option to skip checking layers when comparing output with SIT. ### Details: - Currently only suitable for NRMSE and RRMSE mode. ### Tickets: - EISW-148919 --------- Co-authored-by: Artemy Skrebkov <[email protected]>
### Details: - *Implemented OV CPU tests pipeline with riscV architecture via Xuantie platform (fetched frombinaries, builded by @a-sidorova , sources https://github.com/XUANTIE-RV/xuantie-gnu-toolchain)*) - *todo: python support, building job reusing, implementing support python runner (i.e. using wrapping xuantie application instead of gtests directly)* Implementation of toolchain build - #28142 ### Tickets: - *146964* --------- Co-authored-by: Mikhail Ryzhov <[email protected]>
### Details: - Applied OPENVINO_MATCHER_PASS_RTTI helper macro in Transformations (except in op_convertions - done in #28150). ### Tickets: - CVS-159567 --------- Signed-off-by: Tomasz Jankowski <[email protected]> Co-authored-by: Michal Lukaszewski <[email protected]>
### Details: - Added RTTI definition for passes derived from `ov::pass::GraphRewrite`. - Applied where applicable. ### Tickets: - CVS-159694 Signed-off-by: Tomasz Jankowski <[email protected]>
### Details: - See microsoft/vcpkg#42259 (comment)
### Details: - This issue is a MSVC compiler bug affecting certain versions of Visual Studio 2022. When using std::mutex a null dereference may occur, leading to a silent crash in Release mode. - Adding the compiler option "/D_DISABLE_CONSTEXPR_MUTEX_CONSTRUCTOR" serves as a workaround for this problem. Reference: https://hydrogenaud.io/index.php/topic,126070.0.html https://github.com/microsoft/STL/wiki/Changelog#vs-2022-1710 ### Tickets: - CVS-159684
### Details: - *Skip reorder opt when its dependency is crop* ### Tickets: - *155068*
### Details: - Fix ”Assertion `static_impl != nullptr' failed“ issue for reorder (Convolution_0_reorder_1) node by skipping mark node in shape flow. ### Tickets: - [CVS-154336](https://jira.devtools.intel.com/browse/CVS-154336) --------- Signed-off-by: yuan.xiong <[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.
No description provided.