AI Verify Test Engine

Description

AI Verify Test Engine provides core interfaces, converters, data, model and plugin managers to facilitate the development of tests for AI systems. It is used as a base library for all AI Verify official stock-plugins and can be used to develop custom plugins.

Installation

Install aiverify-test-engine via pip. The following table list the available install options and the optional dependencies along with the additional functionality that is supported.

Developer Guide

Local Installation

To contribute changes to the test engine code, clone the repository, navigate to the aiverify-test-engine folder, and install the dev version of the library:

pip install '.[dev]'

Here's an overview of the project folder structure and a brief description of each:

aiverify-test-engine/
├── aiverify_test_engine/   # Core library code
│   ├── interfaces/         # Core interfaces (algorithm, converter, data, model, pipeline, serializer, plugin)
│   ├── io/                 # Data and model IO related logic
│   ├── plugins/            # Manage the loading and execution of algorithm, data, model, pipeline and plugins
│   ├── utils/              # Utility functions and validators
├── tests/                  # Test cases
├── pyproject.toml          # Project configuration file

Running Tests

python -m pytest tests

If Tensorflow has trouble with GPU, use following environment variable to force Tensorflow to use CPU:

CUDA_VISIBLE_DEVICES="" TF_CPP_MIN_LOG_LEVEL=2 TF_NUM_INTEROP_THREADS=1 TF_NUM_INTRAOP_THREADS=1 XLA_FLAGS="--xla_gpu_unsafe_fallback_to_driver_on_ptxas_not_found" python -m pytest tests

Building the Package

hatch build

License

• Licensed under Apache Software License 2.0

Developers:

• AI Verify