Alina Kladieva
fda3f5d237
* [requirements] Set TF to 2.5.0 * [ngraph|paddlepaddle|requirements] Gast to 0.4.0 (required for TensorFlow 2.5) * [ stress_tests/scripts/requirements] Upd versions * [time_tests/test_runner/requirements] Upd versions * [ngraph|paddle|requirements] Paddle strict major * Gast version to 0.3.3 * Update requirements_dev.txt
Time Tests
This test suite contains pipelines, which are executables. The pipelines measure the time of their execution, both total and partial. A Python runner calls the pipelines and calcuates the average execution time.
Prerequisites
To build the time tests, you need to have OpenVINO™ installed or build from source.
Measure Time
To build and run the tests, open a terminal, set OpenVINO™ environment and run the commands below:
- Build tests:
mkdir build && cd build
cmake .. && make time_tests
If you don't have OpenVINO™ installed you need to have the build folder, which
is created when you configure and build OpenVINO™ from sources:
cmake .. -DInferenceEngine_DIR=$(realpath ../../../build) && make time_tests
For old versions of OpenVINO™ from sources use -DInferenceEngineDeveloperPackage_DIR:
cmake .. -DInferenceEngineDeveloperPackage_DIR=$(realpath ../../../build) && make time_tests
- Run test:
./scripts/run_timetest.py ../../bin/intel64/Release/timetest_infer -m model.xml -d CPU
- Run several configurations using
pytest:
pytest ./test_runner/test_timetest.py --exe ../../bin/intel64/Release/timetest_infer
# For parse_stat testing:
pytest ./scripts/run_timetest.py