So, what kind of plebeian uses Tensorflow rather than PyTorch?

So, what kind of plebeian uses Tensorflow rather than PyTorch?

Other urls found in this thread:

en.wikipedia.org/wiki/No-go_theorem#Examples
pytorch.org/docs/master/notes/multiprocessing.html
twitter.com/SFWRedditVideos

I don't know, but you don't use either.

Prove it.

Tensorflow is static, whereas PyTorch is dynamic, which makes it instantly superior.

I can't prove a negative, but the onus is on you to show you have done enough to have an informed opinion on this. Your post is one provocative sentence without any facts marshaled.

Are you talking about PyTorch being written entirely in Python? There are advantages and disadvantages of that. One effect is you can't introduce any multithreading to speak of. As far as I know, TensorFlow was designed with compute clusters and mobile devices in mind.

Do you mean the data structures can change? It's true that you can have the data structures evolve in time in PyTorch, but this is functionality really used all that often?

>Are you talking about PyTorch being written entirely in Python?
No, I'm talking about static vs dynamic computational graphs.

>It's true that you can have the data structures evolve in time in PyTorch, but this is functionality really used all that often?
Yes.

Also superior syntax. PyTorch master race. TFaggots BTFO.

t. guy who just started watching the FastAI videos

Try doing this is in Tensorflow, tfaggots. Oh wait you can't.

t. projecting

T.Faggot

Okay. Show me what you've built with this functionality.

>I can't prove a negative
>what is Euclid's proof
>what is Fermat's last theorem
>what is math

To be fair, he didn't say YOU couldn't prove a negative, just that he couldn't.

> conflating a proof by contradiction in mathematics with a challenge to show a situation in the physical world has never happened
> doesn't understand Wittgenstein, Russel or Popper but accuses me of not knowing any math

en.wikipedia.org/wiki/No-go_theorem#Examples

I used sci-kit at uni and it was bretty good. Is pytorch better?

> continuing to be this obtuse, as if anyone is going to produce a mathematical proof that a particular unnamed stranger on the internet hasn't programmed, and the point wasn't that OP's post should be backed up experience

>you don't use either
>I can't prove a negative
You aren't human. Don't ask me to justify why. You see, I can't prove a negative.

My god that cake looks delicious

It will be $50 plus tip.

For one slice.

scikit is for general machine learning (think k-means, SVMs, random forests, etc), Torch is for deep learning and neural nets in particular.

PyTorch supports multiprocessing
pytorch.org/docs/master/notes/multiprocessing.html

I never challenged you to prove a situation never came to pass (and no, not a situation like information escaping from a black hole--a situation like someone giving an opinion without explanation).

As my second post stated, I cannot definitely prove you've never used either system (you're welcome to suggest a strategy for this; it would be impressive), but when someone offers strong opinions that don't back up, my money is on their opinions not being based on experience.

>pytorch.org/docs/master/notes/multiprocessing.html
Huh. That's nice. I didn't know that.

Holy shit, is there anything more embarrassing than a retarded faggot using big words to sound smart ?

...

Maybe one with opinions that they can't give any reasons for. Dunno though.

> explanation
> situation
> "came to pass"
> BIG WORDS

Hell no. What is this, delivery to Antarctica?

The cake was baked in the Black Forest, then shipped to a confectionery Cambridge, MA for frosting while still warm. The $50 bill in part covers the supersonic fueling fee.

I understand your misgivings about the cost, but you get what you pay for, y'know?

>Using neural networks framework
>Calling others plebeian

>python
>superior
pick one

Which one is more pythonic?