Lesson 8 (2019) discussion & wiki

Part 2 (2019)
60

Jeremy is answering it :wink:

2 Likes
61

Really excited about this new approach to teaching part 2, the ability to dig deep into the code and develop skillsets that will be transfer to any project is very useful.

1 Like
63

Ostensibly: https://github.com/tensorflow/swift/blob/master/docs/WhySwiftForTensorFlow.md

Realistically: Because Chris Lattner is the one leading the charge and Chris Lattner created Swift :slight_smile:

I’m a little concerned because there isn’t really a Swift story for Windows at the moment.

5 Likes
64

How far is Swift from being ready to be used for Deep Learning as easily as we can use Python now ? 6 months ? 1 year ? 4 years ?

5 Likes
65

Any guide to setup Swift on Windows 10/7 ? Would be a great help !

3 Likes
66

Not sure how I feel about Swift because of the ecosystem. I was very excited about Julia a few years ago, but it seems like never took off. I am excited to learn though.

3 Likes
67

Since we all love PyTorch so much, is there any chance of Swift for PyTorch in the future?

5 Likes
68

Let it be 6 months please

1 Like
69

Dedicating 2 classes for S4TF doesn’t seems like a good investment of time. It’s years behind Pytorch and tensorflow. Tensorlfow 2 seems much more promising. I’m very confused about this move for fast.ai. IS fast,ai going to be halted in term of python development? there is so much left to do


6 Likes
70

This is the guide they link to in Swift for TensorFlow repository: https://swiftforwindows.github.io/

I don’t think it’s ready for primetime. It’s a repository created by a single person and hasn’t had a commit since October.

1 Like
71

Are we still getting an overview of deep RL in part2?

1 Like
72

Jeremy isn’t really convinced by RL I think, at least for now :wink:

73

Swift is a great general-purpose language, I would say. With very powerful generics system, performance, and very modern and convenient syntax. Also, there is lots of similarity with Python’s syntax, and some borrowings from functional languages as well.

2 Likes
74

Swift looks like scala in terms of type safety. On the other hand, the Java Virtual
Machine is heavy. Would it be a good idea to look into Scala for deep learning?

1 Like
75

Please read the post fast.ai Embracing Swift for Deep Learning for more motivation and details:

"It is very early days for Swift for TensorFlow. We definitely don’t recommend anyone tries to switch all their deep learning projects over to Swift just yet! Right now, most things don’t work. Most plans haven’t even been started. For many, this is a good reason to skip the project entirely.

“But for me, it’s a reason to jump in! I love getting involved in the earliest days of projects that I’m confident will be successful, and helping our community to get involved too. Indeed, that’s what we did with PyTorch, including it in our course within a few weeks of its first pre-release version. People who are involved early in a project like this can have a big influence on its development, and soon enough they find themselves the “insiders” in something that’s getting big and popular!”

17 Likes
77

What should we do in the meantime if we are interested in the ‘swift for tensorflow’ between now and 5 weeks from now, now that we’ve been teased all this python is soon to be deprecated :)?

78

RL has never been a part of part 2.

79

And also there are studies like this one out there: https://arxiv.org/abs/1803.07055

2 Likes
80

Thanks for helping. I observed the same . the latest build from that repo is Swift 2.0 while the current stable release is Swift 4.X i guess :frowning:

81

Is numpy a “non-data science module”?

2 Likes