This is a Python library that utilizes GPU acceleration for numerical computing, which is compatible with NumPy and SciPy. You can easily migrate your existing NumPy/SciPy code to run on NVIDIA CUDA or AMD ROCm platforms, with some cases seeing a speed increase of over 100 times.