I started my deep learning journey circa 2018 hand crafting autoencoders with Matlab cells. Pretty quickly moving on to Python, I remember reading about PyTorch and TensorFlow. Back then, TF lacked dynamic graphs which made PT way more favorable. Even still, many compared PT to Keras. However I always thought a closer comparison to Keras was something like Fast.ai from Jeremy Howard et. al., which I really enjoyed and don’t see many people use.

In any case, I ended up sticking with PT and am extremely grateful for all the work put into it. Thank you.