Tensorflow to Pytorch User

I recently switched to Pytorch because of the flexibility and all the research/work that has been done with it. I am finding it difficult to get a hang of, at least for custom loss functions, and all of the datatypes/arrays and how to convert/manipulate them. What are some goods tips/tricks or quick videos/courses that can can help me solve these issues?

Hi Elad!

In addition, of course, to studying the pytorch api documentation, you
should take a look at the pytorch tutorials, many of which are good.

You could do worse than start with pytorch’s 60 Minute Blitz tutorial.

The kinds of documentation and tutorials that work well for someone
is largely a matter of personal taste, but its worth checking out those
provided by the pytorch project.

Pytorch and tensorflow have similar higher-level concepts and goals,
but differ in their detailed implementations. So I would recommend
trying to forget the details of tensorflow, as pytorch will likely be different
(but still make use of the concepts you learned working with tensorflow).

And if you have questions after you’ve tried something specific, post
some (complete, runnable) sample code here, and ask questions.


K. Frank