This small project aims to setup minimal requirements in order to run PyTorch computatiuons on AMD Radeon GPUs on Windows 10 and 11 PCs as natively as possible. It utilizes ZLUDA and AMD's HIP SDK to make PyTorch execute code for CUDA device on AMD, with near native performance. Python module can be run directly on Windows, no WSL needed.