Hacker News Board Hacker News, browsed like a classic message board
Live data from Hacker News

Tensorflow 2.0 Beta 0

github.com

Post reply on HN
Print view View original

31–40 of 44 posts

Re: Tensorflow 2.0 Beta 0

#31

Post by m0zg »

jeffshek wrote: post #7
forgotmyhnacc wrote: post #6

Earlier quoted context omitted.

What are you looking for in deployment friendliness? There's TorchScript to run your code faster (which is a work in progress)

One of the major benefits of TF 2.0 is apparently the capability to quickly deploy to TPU units with a single parameter change. (I haven't tried it, just followed the marketing). AFAIK, This is still being worked on PyTorch via XLA, but not quite there yet.

I've found that with TF in general you can only go "quickly" if everything works. If anything is busted you're more or less screwed because it's so opaque. In contrast, PyTorch lets you inspect whatever you want by setting a pdb breakpoint, and when it gives you errors you most of the time can figure out what's wrong without debugging. The importance of this cannot be overstated.

Re: Tensorflow 2.0 Beta 0

#32

Post by panpanna »

favorited wrote: post #24
panpanna wrote: post #21

Earlier quoted context omitted.

From what I understand this is mostly because they hired the Swift guy. I understand the benefits compared to Python (although I would have preferred Go or Kotlin). But what happens when the guy eventually moves on in a year or two?

Google's making a big investment in Swift, so if Chris left and they were interested in continuing to support it they shouldn't have a problem. I've gone to Swift on the Server conferences hosted/sponsored by Google, their (non-TF) Swift teams are building some cool Swift tools, etc.

Google is a Go shop, specially when it comes to servers.

Swift is not properly supported on linux, which is Google's main platforms.

Re: Tensorflow 2.0 Beta 0

#33

Post by favorited »

panpanna wrote: post #32
favorited wrote: post #24

Earlier quoted context omitted.

Google's making a big investment in Swift, so if Chris left and they were interested in continuing to support it they shouldn't have a problem. I've gone to Swift on the Server conferences hosted/sponsored by Google, their (non-TF) Swift teams are building some cool Swift tools, etc.

Google is a Go shop, specially when it comes to servers. Swift is not properly supported on linux, which is Google's main platforms.

Google is not at all a Go shop. Google is a C++ shop, especially when it comes to servers.

Re: Tensorflow 2.0 Beta 0

#34

Post by bge0 »

panpanna wrote: post #32
favorited wrote: post #24

Earlier quoted context omitted.

Google's making a big investment in Swift, so if Chris left and they were interested in continuing to support it they shouldn't have a problem. I've gone to Swift on the Server conferences hosted/sponsored by Google, their (non-TF) Swift teams are building some cool Swift tools, etc.

Google is a Go shop, specially when it comes to servers. Swift is not properly supported on linux, which is Google's main platforms.

This is incorrect; swift is open source and has linux deployments : https://swift.org/download/

Re: Tensorflow 2.0 Beta 0

#35

Post by panpanna »

favorited wrote: post #33
panpanna wrote: post #32

Earlier quoted context omitted.

Google is a Go shop, specially when it comes to servers. Swift is not properly supported on linux, which is Google's main platforms.

Google is not at all a Go shop. Google is a C++ shop, especially when it comes to servers.

(that contradicts your earlier comment)

Google has been gradually moving away from C++ and Java since 2012. See this quora post with multiple references from Google employees.

https://www.quora.com/How-is-Go-used-at-Google-What-could-be...

Re: Tensorflow 2.0 Beta 0

#36

Post by panpanna »

bge0 wrote: post #34
panpanna wrote: post #32

Earlier quoted context omitted.

Google is a Go shop, specially when it comes to servers. Swift is not properly supported on linux, which is Google's main platforms.

This is incorrect; swift is open source and has linux deployments : https://swift.org/download/

Linux support is still WIP. Some parts are missing

Re: Tensorflow 2.0 Beta 0

#37

Post by pjmlp »

panpanna wrote: post #32
favorited wrote: post #24

Earlier quoted context omitted.

Google's making a big investment in Swift, so if Chris left and they were interested in continuing to support it they shouldn't have a problem. I've gone to Swift on the Server conferences hosted/sponsored by Google, their (non-TF) Swift teams are building some cool Swift tools, etc.

Google is a Go shop, specially when it comes to servers. Swift is not properly supported on linux, which is Google's main platforms.

Google is a more of C++/LLVM, Java/Kotlin, Python shop, than Go one.

You will even notice that it is seldom supported when they announce new server products SDKs.

Re: Tensorflow 2.0 Beta 0

#38

Post by pjmlp »

panpanna wrote: post #35
favorited wrote: post #33

Earlier quoted context omitted.

Google is not at all a Go shop. Google is a C++ shop, especially when it comes to servers.

(that contradicts your earlier comment) Google has been gradually moving away from C++ and Java since 2012. See this quora post with multiple references from Google employees. https://www.quora.com/How-is-Go-used-at-Google-What-could-be...

That is not what their employees talk about at CppCon, LLVM conferences, ISO C++ meetings, Java Language Summit (yes they come around in spite of Android).

Go is mostly a Docker/Kubernetes thing.

Re: Tensorflow 2.0 Beta 0

#39

Post by credit_guy »

p1esk wrote: post #28
credit_guy wrote: post #27

Earlier quoted context omitted.

Maybe I'm not up to speed with the latest PyTorch, but to me Keras feels much more natural. In Keras if you want to define a deep learning network, then you just do that, you specify the first layer, the second layer, etc, then you calibrate over some test and validation samples, using a certain flavor of gradient descent, for a given loss function. In PyTorch, you have to define a class, with a constructor, some met…

in Keras I can't (easily) change the architecture of a learner after I defined it I'm not sure what you mean here, because only PT lets you change architecture after you define it, while TF/Keras uses a static precompiled graph. Now that's changing with eager mode, but that used to be the main advantage of PT.

That was what I heard too, that in PyTorch you could adaptively build your graph. In Keras you could in principle emulate this in a brute-force way: you define the graph, train it, saving the weights, analyze them, decide what nodes to remove and what nodes to add, and then rebuild the graphs from scratch and use the adjusted weights from the previous round. I say in principle, personally I never did it. Some people hint that in PT this is a breeze, I'd be curious to see some example. If you have any links, that would be much, much appreciated.

Re: Tensorflow 2.0 Beta 0

#40

Post by mark_l_watson »

flensortow wrote: post #15

Any news on Swift for Tensorflow? I’m skeptical of how much practical benefit it will provide but still willing to take a look at it. There doesn’t seem to be any mention of it here.

I have spent a few evenings playing with early releases. I like the ‘turtles all the way down’ idea, but I am waiting to see more mature releases. I have spent much more time with TensirFlowJS that works well, has many great examples.
Post reply on HN