I first started working with tensorflow 1.x as a backend to keras before they merged. With cuda installed to take advantage of GPU acceleration, my python code ran pretty snappy
