DimTorch

A light wieght Deep Learning Framework for Learning Purposes.

Tested With

• CMake (version 3.28.3)
• g++ (Ubuntu 13.3.0-6ubuntu2~24.04.1) 13.3.0
• Linux Mint 22.3 with Kernal 6.17.0-42-generic
• Libtorch with build number 2.13.0+cpu (This is only needed by the unit tests to check the aganst DimTorch)
• CPU: i5-10400 desktop (supports AVX2, not AVX512)

Building

This project uses CMake. To build:

mkdir build
cd build
cmake ..
make

Running Tests

After building, run the unit tests from the build directory:

./unit_tests # will run some basic tests aganst libtorch. They are a bit limited so the plan will be to test module by module in the future
./mnist_test # will train a basic model on the mnist data test
./mm_test    # will benchmark the matrix multiplication operation on two 4800 by 4800