Recently, I was interested in Deep Learning / Machine Learning / Neural Networks / ….
A quick look at the literature makes me believe the following:
a] If I want to do ML, the preferred choice is either Python or C. C is usually faster, but the syntax is more complicated. Python is perfect for playing around, but is also used by professionals. This is due to powerful libraries like TensorFlow. C code can be used in Python using wrappers. A famous example of this is NumPy / Pandas.
b] The training takes place on the GPU and usually CUDA is used for this. Unfortunately, CUDA is limited to NVIDIA GPUs.
c] Training on a non-NVIDIA GPU uses OpenCL.
My situation is that I'm using a pretty powerful Mac with a 4GB GPU and this seems to be sufficient for starters. Unfortunately, it comes with a non-NVIDIA GPU and so I have to use OpenCL. On the other hand, I would like to use TensorFlow, for example, and work with Python.
The only thing that made a quick search on the Internet possible was this Python wrapper, which unfortunately seems to work for Windows and Linux, but not for Mac OS.
Here is my question: How can I train on the GPU of my Mac and use Python? With that I mean, Do the programming in Python and not Write the code in C and use a Python wrapper,