150 points by ml_comparison 6 months ago flag hide 10 comments
deeplearning_king 6 months ago next
Great comparison! I've been using TensorFlow for a while but PyTorch's simplicity amazes me.
tensorflow_lover 6 months ago next
@deeplearning_king while PyTorch has a nice syntax I think TensorFlow has better performance for large scale projects.
pytorch_champ 6 months ago next
@tensorflow_lover it may be true but you can't ignore the fact that for research purposes PyTorch is better. What do you think?
king_of_nets 6 months ago next
@tensorflow_lover true but with the release of TensorFlow 2.0 the simplicity of TF is quite similar to PyTorch. What's your thoughts on this?
user1234 6 months ago prev next
I think it depends on the goal. If you're trying to win a specific machine learning competition TensorFlow might be better, but for fast prototyping and research, PyTorch is the way to go.
ella_ai_girl 6 months ago next
@user1234 I completely agree with your point. Given a choice I'd always prefer PyTorch's dynamic computational graph vs. TensorFlow's static graph.
tensorflow_lover 6 months ago next
@ella_ai_girl yeah, even I initially had problems adjusting to TF's static graph but with 2.0 it's much better now.
neural_networks_rocks 6 months ago prev next
I used to be a TensorFlow fanboy but time and again I've observed members of my team getting more done in less time with PyTorch. Any thoughts on that?
code_ninja_13 6 months ago prev next
I've used both TensorFlow and PyTorch. I prefer PyTorch because of its ease of use and dynamic computational graph. But, both have their place depending on the scenario.
coder_girl 6 months ago prev next
What do people think about the new release PyTorch 1.5 and TensorFlow 2.3? Any major improvements worthy of noting down?