natas_hw
/
openvino
0
0
Fork 0
Code Issues Pull Requests Packages Projects Releases Wiki Activity

[DOCS][Layer Tests] Update layer test documentation (#23485) 

**Details:** Update layer test documentation. It will be used by
external developers and mentioned in GFIs.

**Ticket:** TBD

---------

Signed-off-by: Kazantsev, Roman <roman.kazantsev@intel.com>
Co-authored-by: Anastasiia Pnevskaia <anastasiia.pnevskaia@intel.com>
This commit is contained in:
Roman Kazantsev 2024-03-15 20:17:15 +04:00 committed by GitHub
parent 13df2c4f3a
commit 217afeecb3
No known key found for this signature in database
GPG Key ID: B5690EEEBB952194
1 changed files with 39 additions and 26 deletions
Show Stats Download Patch File Download Diff File Expand all files Collapse all files

65
tests/layer_tests/README.md
View File

@ -1,34 +1,47 @@
# Layer tests
# Layer Tests
This folder layer tests framework code and test files.
The layer tests primarily aim to validate support for PyTorch, TensorFlow, TensorFlow Lite, and ONNX frameworks' operations by OpenVINO.
The test pipeline includes the following steps:
1. Creation of a model with the tested operation using original framework API
2. Conversion of the created model using OpenVINO's `convert_model` method
3. Inference of both the original and converted models using the framework and OpenVINO on random input data
4. Checking whether the inference results from OpenVINO and the framework are the same or different within a tolerance threshold
## Getting Started
## Setup Environment
#### Pre-requisites
* OpenVINO should be configured as usual.
#### Setup
* Install requirements:
```bash
pip3 install -r requirements.txt
To set up the environment for launching layer tests, perform the following steps:
1. Install the OpenVINO wheel package. If you're testing changes in OpenVINO, build your local wheel package for installation.
Find instructions on how to build on [wiki page](https://github.com/openvinotoolkit/openvino/blob/master/docs/dev/build.md).
```sh
pip install openvino.whl
```
* Set up environment variables for layer tests (if you use wheel package path to python api could be removed):
```bash
export PYTHONPATH="path_to_openvino"/tests/layer_tests/:"path_to_openvino"/tools/mo:"path to python api"
2. (Optional) Install the OpenVINO Tokenizers wheel package if you're testing the support of operations using conversion and operation extensions from OpenVINO Tokenizers.
If you're testing changes in OpenVINO Tokenizers, build your local wheel package for installation.
Find instructions on how to build on [GitHub page](https://github.com/openvinotoolkit/openvino_tokenizers)
```sh
pip install openvino_tokenizers.whl
```
* If there is need to use specific libs it is possible to specify path to them using OV_LIBRARY_PATH env variable
```bash
export OV_LIBRARY_PATH="path_to_libs"
3. Install requirements for running layer tests.
```sh
cd tests/layer_tests
pip install -r requirements.txt
```
* To parametrize tests by device and precision (optional)
```bash
export TEST_DEVICE="CPU;GPU"
export TEST_PRECISION="FP32;FP16"
## Run Tests
Set environment variables `TEST_DEVICE` and `TEST_PRECISION` to select device and inference precision for OpenVINO inference. Allowed values for `TEST_DEVICE` are `CPU` and `GPU`. Allowed values for `TEST_PRECISION` are `FP32` and `FP16`.
Example to run the TensorFlow layer test for the `tf.raw_ops.Unique` operation on CPU with default inference precision for device:
```sh
cd tests/layer_tests
export TEST_DEVICE="CPU"
pytest tensorflow_tests/test_tf_Unique.py
```
## Run tests
```bash
py.test
```
Example to run the PyTorch layer test for the `torch.linalg.cross` operation on CPU and GPU with `FP16` and `FP32` inference precisions:
```sh
cd tests/layer_tests
export TEST_DEVICE="CPU;GPU"
export TEST_PRECISION="FP32;FP16"
pytest pytorch_tests/test_cross.py
```