jph00

47 ms·

571.
by jph00 7y ago
Their model didn't win. IBM's model won, based on actual metrics around useful insights.
572.
by jph00 7y ago
It's feature complete now, and we're writing the paper and book about it. So it's ready for early adopters to start prototyping with. Lots of folks already doing so and discussing here: https://forums.fast.ai/
573.
by jph00 7y ago
> Also the distinction between the code & notebooks are too big, it's clunky to move between. You get to resort copy-pasting back-and-forth. We have solved this problem by providing 2-way sync between notebooks and libraries,
574.
by jph00 7y ago
> to pretend that all human/computer interaction fits that mold is to be willfully ignorant of the reality of personal computing He doesn't. He has a whole section where he points out types of software that mice are great for,
575.
by jph00 7y ago
Oh interesting idea - not something I'd thought of. (We don't use Airflow at fast.ai so this will likely require a PR.)
576.
by jph00 7y ago
Hi folks - Sylvain Gugger and I created nbdev. Let me know if you have any questions about it. Since nbdev is written in nbdev itself, you can also use its repo as an example project to see how it works. Another good example project written
577.
Nbdev: Use Notebooks for Everything(fast.ai)
28 points by jph00 7y ago|5 comments
578.
by jph00 7y ago
This is great! It's also very similar to the approach we use to teach deep learning at fast.ai, based on research into teaching methods done by David Perkins at Harvard. A couple of our principals are: - Teaching “the whole game”–start
579.
by jph00 7y ago
Thanks! I haven't been involved involved with Fastmail for the last 10 years, so I can take no credit for how amazing they've been throughout that time - but I'm thrilled to see that they're still doing well. I'm st
580.
by jph00 7y ago
You may find this article I wrote useful background for understanding the capabilities of numerical computing in Swift, including details about a library which provides features overlapping with the new Swift package: https://www
581.
by jph00 7y ago
Interesting! Did you have to make any substantive changes to the ULMFiT approach to make it work for this problem? Did you use the fastai implementation, or write something from scratch? (Disclaimer: I'm a co-author on the ULMFiT paper
582.
by jph00 7y ago
You're spot on - the calculations in the post are totally wrong. For instance, a Dice coefficient is itself the average of hundreds of thousands of observations (each pixel in an image), so you can't just treat the coefficient as
583.
Millions of Americans’ Medical Images and Data Are Available on the Internet(propublica.org)
86 points by jph00 7y ago|33 comments
584.
by jph00 7y ago
That repo is written in Swift, hence it is called Swift Core ML Transformers.
585.
by jph00 7y ago
For those interested in learning more about this, Chris Lattner (project lead on MLIR and Swift for TensorFlow) and I co-taught two lessons covering these topics. They're lesson 13 and 14 here: https://course.fast.ai/vi
586.
Creating a Pop Music Generator with the Transformer(towardsdatascience.com)
5 points by jph00 7y ago|0 comments
587.
by jph00 7y ago
fastai uses held-out data for reporting accuracy by default, which was done in this article.
588.
by jph00 7y ago
> Turning a little more complex elementwise computation into vectorized numpy calls produces a lot of memory overhead. For each vectorized numpy call, which does a basic computation (like adding two operands, or negating, or sinus of, ..
589.
by jph00 7y ago
I used to be a big fan of a similar tool https://www.sococo.com/ . However it seems to have stopped getting much love in recent years. Seems like Tandem is cheaper and has some nice extra features - so looking forward to tr
590.
by jph00 7y ago
Also, note that Greg's goal was to contribute to OpenAI's flagship project. That's a rather ambitious goal! Also, most folks I know that are making practical deep learning contributions are doing so by combining their pre-exi
591.
by jph00 7y ago
You might find it helpful to see what other people are doing with deep learning - see if any of it seems relevant to your interests. For instance, here's some examples of folks from diverse backgrounds using DL in their domains of inte
592.
by jph00 7y ago
The goal is to allow Swift to be used for writing MLIR and XLA kernels. The new LazyTensor under development already allows for fused XLA operations to be created in Swift. There's an awful lot you can do in Swift which is very very ha
593.
by jph00 7y ago
Not in the short/medium term, at least. The stuff we covered in that Intro to ML course hasn't really changed, so the current course is nearly equally relevant today. (We used the fastai library a little for some basic utility fun
594.
by jph00 7y ago
Nice to see this on the front page! :D There's some more info about the Swift lessons on the TensorFlow blog here: https://medium.com/@tensorflow/3ee7dfb68387 Happy to answer any questions you all might have.
595.
Deep Learning from the Foundations(fast.ai)
10 points by jph00 7y ago|0 comments
596.
A LaTeX Add-In for PowerPoint(fast.ai)
3 points by jph00 7y ago|0 comments
597.
by jph00 7y ago
It's a fast moving field. In the past 5 years we've scene enthusiasm for Caffe, luatorch, theano, tensorflow, and most recently pytorch. And tensorflow 2 of course, which totally changes the programming model. Our own library chan
598.
by jph00 7y ago
The fastai library uses a style guide specifically designed for data science libraries: https://docs.fast.ai/dev/style.html . There is only one global in the library, which is a singleton called `defaults`, and contain
599.
by jph00 7y ago
Surprised (and delighted!) to see this pop up on the front page of HN today! This course is designed to teach proficient coders the practice of deep learning, that is: how to train accurate models; how to test and debug models; and key de
600.
by jph00 7y ago
You might also be interested in the sister course to the one posted here: "Introduction to Machine Learning for Coders": http://course18.fast.ai/ml

More ›