Standard APIs for AI operations and automation testing framework.

Environment Requirement

Dependency	Version required
CXX	17 or higher
CMake	3.18 or higher
Gtest	1.8.x or higher
Libconfig	1.7.3 or higher
Libtorch	latest

Project Layout

AutoTest is an operator standard automation testing framework.

• Autotest/auto_test contains comparative test files.
• Autotest/src contains data type and kernels for aitisa.
• Autotest/hice-master contains data type and kernels for hice.
• Autotest/test contains test files for aitisa kernels.
• Autotest/generate contains comparison results.
• Autotest/Config contains some test config.

Here is a comparative test result of the 'activate' operator among the three different operator libraries: Aitisa, Hice, and Torch.

Quickstart

You can setup Standard APIs for AI operations by following instructions:

1. Use git clone instruction to download source code

   git https://github.com/JasonChen9/AutoTest.git

2. Make a new directory build under the project directory, then use cmake instruction

   mkdir build  
   cd build  
   cmake ..

3. Use make instruction to compile the code

   make

4. Run testing file. Take running convolution operator testing file as an example

   cd bin
   ./conv_test

5. Run hice_aitisa_test file to compare hice with aitisa.

    cd bin
    ./hice_aitisa_test

If you want compare with pytorch, you should switch AITISA_API_PYTORCH ON in CMakeList.txt

    option(AITISA_API_PYTORCH "test with pytorch" ON)

Due to the dependency on hice, we are providing installation instructions for both MKL and DNNL libraries:

MKL

https://www.intel.com/content/www/us/en/develop/documentation/installation-guide-for-intel-oneapi-toolkits-linux/top/installation/install-using-package-managers/apt.html

According to flashlight/flashlight#640. One might meet MKL library not found issue. We can use the method below:

   cd /tmp && wget https://apt.repos.intel.com/intel-gpg-keys/GPG-PUB-KEY-INTEL-SW-PRODUCTS-2019.PUB && \
   apt-key add GPG-PUB-KEY-INTEL-SW-PRODUCTS-2019.PUB && \
   sh -c 'echo deb https://apt.repos.intel.com/mkl all main > /etc/apt/sources.list.d/intel-mkl.list' && \
   apt-get update && \
   apt install intel-mkl-64bit-2020.4-912

dnnl

https://www.intel.com/content/www/us/en/develop/documentation/installation-guide-for-intel-oneapi-toolkits-linux/top/installation/install-using-package-managers/apt.html