This repository reproduces the results of arXiv:2212.01175, but implemented in PyTorch with support for GPU parallel computing.
The method uses random walk search with flipping and reduction operations to find optimal decomposition schemes for matrix multiplication tensors. It successfully finds optimal multiplication schemes for matrices of size n × n for n = 2, 3, 4.
Example notebooks demonstrating the method can be found in the notebooks/ directory.