On: The State of Machine Learning Frameworks

[qwerty456127]

Why?

[chillee → qwerty456127]

Well, you could read the article :^)

As a summary, though:

PyTorch has become dominant in research because of its API (both its stability + having eager mode).

TF has become dominant in industry because A. it came out several years before PyTorch and industry is slow to move, B. It supported a lot of production use cases (mobile, serving, removing Python overhead) that PyTorch didn't for a long time.

[qwerty456127 → chillee]

If only articles were as concise as your summary is I would enjoy reading them but as long as they are many pages long I have no time to read beyond the titles, abstracts, conclusions and comments.

BTW I've also read (here on HN) PyTorch learns much faster than TensorFlow does.

[tecleandor → qwerty456127]

Then it wouldn't be an article, it would be a summary :)

[enriquto → chillee]

well, for one pytorch added the python overhead over an otherwise very clean lua jit

[mruts → enriquto]

PyTorch wraps THNN, not Torch. Moreover, even if this was true, it wouldn't matter at all. Practically 0% of the overhead is related to Python in the first place, all of the time is dominated by the underlying C implementation.

[nknealk → qwerty456127]

100% anecdotal --

The Keras interface for tensorflow makes it easy & fast to make "good enough" models. That is often a driving factor

[mruts → nknealk]

I've personally found pytorch easier to use than Keras. It's undoubtably easier to use than Tensorflow, of course.

[hprotagonist → qwerty456127]

research: the API.

industry: tensorRT.