Hey everyone as we kick start this series I would like to shed some light on why I prefer PyTorch over TensorFlow and why you should too.
As a deep learning practitioner I like (or require) custom architectures for my projects or research papers which can be difficult to build using TensorFlow because it’s implementation is not very friendly (non-pythonic). Hence, building a custom architecture in TensorFlow is time-consuming.
Additionally, unless you are using core TensorFlow you don’t have much control over anything other than the number of layers in a model and their configurations.
PyTorch, on the other hand, gives you complete control over how you want to train your model and what should be the underlying architecture. In fact, it uses OOP for building deep learning model and the training loop is an ACTUALLY python for loop!!
This gives you full control over how the model is trained and provides you with all the optimizations that libraries like TensorFlow provide.
Now that I’ve explained why I prefer PyTorch over TensorFlow we are ready to explore deep learning.