Programming on Python with CuPy does not necessarily involve the use of CUDA-C/C++ within your Python code in order to harness NVIDIA GPU devices. You can implement CuPy for NVIDIA GPUs in a way that looks very similar to a traditional NumPy implementation in Python. CuPy's features include the following:
- High performance with CUDA
- Highly compatible with NumPy
- Easy to install
- Easy to write a custom kernel
- Customizable memory allocator and memory pool
- cuDNN utilities