OpenCL is standard for running parallel programs on heterogeneous systems. CUDA is NVIDIA technology helping with running massively parallel programs on GPU.
PyOpenCL is Python library for accessing OpenCL written by Andreas Klöckner. PyCUDA is Python library for accessing CUDA written by Andreas Klöckner.Git repositories for those projects are available on GitHub:
I also maintain Debian packages for PyOpenCL and PyCUDA. You can find more information about those in other page.