PyTorch 2.0

#1

Post by lairv » Fri, Dec 02, 2022, 4:17 PM UTC

PyTorch 2.0
pytorch.org

Re: PyTorch 2.0

#2

Post by marban » Fri, Dec 02, 2022, 4:19 PM UTC

From what I see, still no 3.11 support — Same for Tensorflow which won't ship before Q1 23.

Re: PyTorch 2.0

#3

Post by joelfried » Fri, Dec 02, 2022, 4:36 PM UTC

marban wrote: post #2

From what I see, still no 3.11 support — Same for Tensorflow which won't ship before Q1 23.

Who is still supporting Windows 3.11?

Re: PyTorch 2.0

#4

Post by sairahul82 » Fri, Dec 02, 2022, 4:40 PM UTC

joelfried wrote: post #3

marban wrote: post #2

From what I see, still no 3.11 support — Same for Tensorflow which won't ship before Q1 23.

Who is still supporting Windows 3.11?

It’s python 3.11 :)

Re: PyTorch 2.0

#5

Post by quietbritishjim » Fri, Dec 02, 2022, 4:47 PM UTC

> Today, we announce torch.compile, a feature that pushes PyTorch performance to new heights and starts the move for parts of PyTorch from C++ back into Python.

I'll admit I don't know enough about PyTorch to know what torch.compile is exactly. But does this means some features of PyTorch will no longer be available in the core C++ library? One of the nice things about PyTorch had been that you could do your training in Python then deploy with a pure C++ application.

Re: PyTorch 2.0

#6

Post by minimaxir » Fri, Dec 02, 2022, 4:58 PM UTC

marban wrote: post #2

From what I see, still no 3.11 support — Same for Tensorflow which won't ship before Q1 23.

It's not a huge deal, as the speed improvements in 3.11 likely wouldn't trickle down at the core PyTorch level.

Re: PyTorch 2.0

#7

Post by belval » Fri, Dec 02, 2022, 5:19 PM UTC

> We believe that this is a substantial new direction for PyTorch – hence we call it 2.0. torch.compile is a fully additive (and optional) feature and hence 2.0 is 100% backward compatible by definition.

How about just calling it PyTorch 1.14 if it's backward compatible? Version numbering shouldn't be used as a marketing gimmick.

Re: PyTorch 2.0

#8

Post by posharma » Fri, Dec 02, 2022, 5:20 PM UTC

belval wrote: post #7

> We believe that this is a substantial new direction for PyTorch – hence we call it 2.0. torch.compile is a fully additive (and optional) feature and hence 2.0 is 100% backward compatible by definition. How about just calling it PyTorch 1.14 if it's backward compatible? Version numbering shouldn't be used as a marketing gimmick.

Is this really the biggest problem that needs to be solved in AI?

Re: PyTorch 2.0

#9

Post by poorman » Fri, Dec 02, 2022, 5:24 PM UTC

Sill no support for/from Apple Silicon?

Re: PyTorch 2.0

#10

Post by robertlagrant » Fri, Dec 02, 2022, 5:26 PM UTC

posharma wrote: post #8

belval wrote: post #7

> We believe that this is a substantial new direction for PyTorch – hence we call it 2.0. torch.compile is a fully additive (and optional) feature and hence 2.0 is 100% backward compatible by definition. How about just calling it PyTorch 1.14 if it's backward compatible? Version numbering shouldn't be used as a marketing gimmick.

Is this really the biggest problem that needs to be solved in AI?

Quite the non sequitur you have there.