TensorFlow vs. PyTorch to start a career

Question

For a person starting out in deep learning, which one do you think is better based on:- ease of learning- job abundance- future dominanceIn my review i saw that most of pytorch comes from students/academics doing POC and putting things together.- philosophically tensoflow seems more appealing and well thought architecture.- in many example codes i saw tf constructs the model/graphs first as compared to a lot of for loops in pytorch examples.- pytorch is claiming to be easy and doing a huge marketing (even ads) and i guess it will attract more new users .What's your take?

davidatbu · Accepted Answer

I can't comment on the dominance of either framework on production-level deployment, but when it comes to academic work, Pytorch is unequivocaly winning. There were some studies showing the number of mentions of deep learning frameworks in the top ML conferences (ICLR, NAACL, ..etc) that demonstrated this. I'd expect that to trickle down to the industry in the coming years.

There's Andrej Karpathy's (chief of AI at Tesla) endorsement too:

    I've been using PyTorch a few months now and I've never felt better. I have more energy. My skin is clearer. My eye sight has improved.

p1esk · Answer

It&rsquo;s like C++ vs Java. Both are good and popular. Which one to learn first does not really matter.Don&rsquo;t waste your time on questions like this. Start learning the concepts- they are the same regardless if framework.