Anyone have a ballpark estimate of Tensorflow models in production vs PyTorch? Is it something like 90/10?
Looks like 3:1 ish.
https://thegradient.pub/state-of-ml-frameworks-2019-pytorch-...
Deep Learning with PyTorch
31–40 of 40 posts
Re: Deep Learning with PyTorch
#32Anyone have a ballpark estimate of Tensorflow models in production vs PyTorch? Is it something like 90/10?
Looks like 3:1 ish. https://thegradient.pub/state-of-ml-frameworks-2019-pytorch-...
Which estimate are you looking at? The main thing I see is:
“PyTorch and TensorFlow for Production
Although PyTorch is now dominant in research, a quick glance at industry shows that TensorFlow is still the dominant framework. For example, based on data from 2018 to 2019, TensorFlow had 1541 new job listings vs. 1437 job listings for PyTorch on public job boards, 3230 new TensorFlow Medium articles vs. 1200 PyTorch, 13.7k new GitHub stars for TensorFlow vs 7.2k for PyTorch, etc.”
That suggests 1:1 for jobs, 2:1 for github stars and 3:1 for articles on Medium. Really hard to say if any of those reflect uses in production in any meaningful way, but if so, I’d suggest that the jobs and/or github stars might be more useful than articles on Medium.
Re: Deep Learning with PyTorch
#33Shouldn't [0,2,1] = 5 on page 17 and not 2? Or do i misunderstand how pytorch handle tensors?
You are correct.
Re: Deep Learning with PyTorch
#34Anyone have a ballpark estimate of Tensorflow models in production vs PyTorch? Is it something like 90/10?
I have used TensorFlow almost exclusively for my deep learning work in the last five years and using TensorFlow is much more common, largely I think, because Andrew Ng’s excellent five part course used it. On the other hand Jerry Howards’s wonderful free classes use mostly PyTorch so maybe my argument does not hold.
For research and for use in published papers, PyTorch is much more widely used.
Re: Deep Learning with PyTorch
#35What kind of knowledge should you have before getting into this book? I've been meaning to try to learn ML and have been looking at some university courses that have all the material available.
Goodfellow’s book on deep learning[0] is a good starter - the first chapters give a solid overview of ML theory as well. Elements of Statistical Learning is another. [0] http://www.deeplearningbook.org/
I retired this year, but my last job was managing a deep learning team. I know many people who own this book, but no one including myself who has read it. Personally, the value I got from it was the first section on math, then picking and choosing limited material that I used for a reference or overview.
Re: Deep Learning with PyTorch
#36Earlier quoted context omitted.
Looks like 3:1 ish. https://thegradient.pub/state-of-ml-frameworks-2019-pytorch-...
Which estimate are you looking at? The main thing I see is: “PyTorch and TensorFlow for Production Although PyTorch is now dominant in research, a quick glance at industry shows that TensorFlow is still the dominant framework. For example, based on data from 2018 to 2019, TensorFlow had 1541 new job listings vs. 1437 job listings for PyTorch on public job boards, 3230 new TensorFlow Medium articles vs. 1200 PyTorch,…
The jobs numbers suggests where the trend is going. The github and article counts seem more useful to indicate the current state. When looking at those numbers I'd rather pick the more conservative number.
Re: Deep Learning with PyTorch
#37To those who don't want to submit their email: https://pytorch.org/assets/deep-learning/Deep-Learning-with-...
Damn, I should read the comments first
Re: Deep Learning with PyTorch
#38Earlier quoted context omitted.
Goodfellow’s book on deep learning[0] is a good starter - the first chapters give a solid overview of ML theory as well. Elements of Statistical Learning is another. [0] http://www.deeplearningbook.org/
I retired this year, but my last job was managing a deep learning team. I know many people who own this book, but no one including myself who has read it. Personally, the value I got from it was the first section on math, then picking and choosing limited material that I used for a reference or overview.
I read it! But then, I love reading technical books.
I did find that it didn't provide much context around why the equations matter, and definitely wouldn't be useful for those starting out in the field. It did have some pretty good coverage of gradient descent and various optimisers, which I found useful.
tl;dr: not really worth it for its stated purpose, but not a bad second or third stats book.
Re: Deep Learning with PyTorch
#39Earlier quoted context omitted.
Goodfellow’s book on deep learning[0] is a good starter - the first chapters give a solid overview of ML theory as well. Elements of Statistical Learning is another. [0] http://www.deeplearningbook.org/
I don’t think it is a good book. From a didactic point of view, I actually found it one of the worst resources out there. The math intro at the beginning is too superficial - either you know it and skip the chapter, or you need another resource to learn. The rest of the first part is okay, but parts 2 and 3 are really not very helpful to someone who doesn’t already understand it. I strongly recommend fast.ai instead.…
I finally started looking into fast.ai and the setup seems to be a lot of hassle. It doesn't help that the course dismissively just says "just buy server time even if you have a GPU for this."
Re: Deep Learning with PyTorch
#40Earlier quoted context omitted.
I don’t think it is a good book. From a didactic point of view, I actually found it one of the worst resources out there. The math intro at the beginning is too superficial - either you know it and skip the chapter, or you need another resource to learn. The rest of the first part is okay, but parts 2 and 3 are really not very helpful to someone who doesn’t already understand it. I strongly recommend fast.ai instead.…
I finally started looking into fast.ai and the setup seems to be a lot of hassle. It doesn't help that the course dismissively just says "just buy server time even if you have a GPU for this."
I feel the same way! My workstation can easily handle the fast.ai projects but they insist that people use cloud servers.