TensorFlow vs PyTorch

neural-network
deep-learning
machine-learning
question

#1

I have been using TensorFlow for a while now. It seems everyone else is moving towards PyTorch because its more dynamic and TensorFlow seems static.

Working with large data sets is extremely time consuming so before we venture and try out PyCharm, I would like to listen to some opinions here. Thanks!


#2

Not everyone! but seems like a lot of people are moving towards PyTorch. But isn’t that the case for any technology? someone has to start and then we create something better or different.

In my opinion, the major difference between TensorFlow and PyTorch is the way computational graphs are represented. Static (TensorFlow) vs. dynamic graph representations (PyTorch)…

In Tensorflow, you have to pre define the computation graph of your model. But in PyTorch, you can define or even manipulate the graph at any time. This is particularly helpful while using variable length inputs in RNNs.

One great benefit of TensorFlow is the boards (TensorBoard). It lets you visualise the ML models within the browser.

It comes down to the taste. TensorFlow still have a much bigger community behind it and who is to say that it wont evolve past this point.

Both TensorFlow and PyTorch have contributed immensely to the community.


#3

Tensorflow is math library of deep learning for complex mathematical operations.
Pytorch gives us liberty to work on pre build models and use them as per our requirements.
Its best models are good for image classification problems.