jph00

28 ms·

391.
by jph00 4y ago
Yes, after we create something from scratch, we then see how to use implementations from libraries such as PyTorch. For basic operations like convolution, whilst we do create a GPU-accelerated version, it's not as fast as PyTorch/
392.
by jph00 4y ago
The math is covered in this course. We look at both ordinary and stochastic differential equations, and at Langevin dynamics / score matching, as well as the needed probability foundations. We only cover the necessary bits of these top
393.
by jph00 4y ago
Hi folks. Nice to see our new free (and ad-free) course here on HN! This course is for folks that are already comfortable training neural nets and understand the basic ideas of SGD, cross-entropy, embeddings, etc. It will help you both unde
394.
by jph00 4y ago
Low-rank adaptation - an efficient way to fine tune some types of model. https://huggingface.co/blog/lora
395.
by jph00 4y ago
A lot of replies here which are based on out of date information and misunderstandings about deep learning super resolution (DLSR). Firstly, this isn't just doing edge detection etc as happened 20 years ago. It's creating a deep l
396.
by jph00 4y ago
Amazing this book is from 2003 - very prescient.
397.
by jph00 4y ago
CTE==Common Table Expression. It's not specific to any particular language. It's basically like a view, except you can define them the same place as you use them, and they can be recursive.
398.
by jph00 4y ago
It's not "throughout the whole article". It looks like he has updated a previous post and made one editing mistake, which he then copied incorrectly once. In all other mentions he correctly describes RTX 40 as Ada. The analys
399.
by jph00 4y ago
A couple of weeks ago a new paper came out that shows how to train a high quality language model on a single GPU in one day. https://arxiv.org/abs/2212.14034
400.
by jph00 4y ago
fast.ai has both - the "part 1" section is top-down, and the "part 2" section is bottom up. You can do part 2 without having done part 1. Part 2 starts with implementing matrix multiplication from scratch, then backprop
401.
by jph00 4y ago
Someone involved with Stability.ai posted this article about the "leak": https://danieljeffries.substack.com/p/why-the-future-of-open...
402.
First Two Lessons of "From Deep Learning Foundations to Stable Diffusion"(fast.ai)
4 points by jph00 4y ago|0 comments
403.
by jph00 4y ago
Actually we sold FastMail to Opera, and since it didn't really work out, bought it back again. Based on that experience, I'd be surprised if FastMail got (re-)acquired...
404.
Practical Deep Learning for Coders(fastai.github.io)
5 points by jph00 4y ago|0 comments
405.
Deep Learning Foundations Signup and Open Source Scholarships(fast.ai)
1 points by jph00 4y ago|0 comments
406.
by jph00 4y ago
Cyrus IMAP has been using skiplists for its storage for around 20 years IIRC. https://www.cyrusimap.org/imap/concepts/deployment/databases... I was very excited to come across this when I was working at fastm
407.
by jph00 4y ago
Awesome! I think a lot of the fast.ai community would be interested to learn from your experiences.
408.
by jph00 4y ago
Was this link meant to have a timestamp? I was wondering where it came from - I found it here: https://youtu.be/htiNBPxcXgo?t=5611 (I know I've shared this factoid in a few interviews and talks, but didn't remembe
409.
by jph00 4y ago
I'd be very interested!
410.
by jph00 4y ago
Yes, but more from a "what do they do, and how do we code them" than "how are they derived mathematically". I think sampling methods might be the bit of this which is the hardest to fully understand given our prerequisit
411.
by jph00 4y ago
Sorry folks for not contributing to this thread earlier - didn't realise this popped up on HN while I was sleeping! I'm Jeremy, and I'll be running this course. Lemme know if you've got any questions about it, or anythin
412.
by jph00 4y ago
I helped curate this list: https://www.modulo.app/all-resources/bestappsforlearning Glad you liked course.fast.ai :D
413.
by jph00 4y ago
We generally found our tutors via modulo.app. The percent I spend personally varies quite a bit. We'll homeschool as long as it seems the best for our child and as long as she wants to.
414.
by jph00 4y ago
Hi all. I'm Jeremy, the husband of the author of this post. We really didn't expect to see this get any particular attention -- but I'm happy to answer questions about homeschooling you might have. As my wife Rachel said in t
415.
by jph00 4y ago
Having not gone through it yourself, you may have an overly-rosy view of school. Personally, I hated school, and I think it had a strongly net negative impact on me. It's not for everyone.
416.
by jph00 4y ago
Here's a single line you can run that'll add the needed line to the file in the correct place for you: sudo perl -pi -e 's/(pam_smartcard.so)/$1\nauth sufficient pam_tid.so/' /etc/pam.d&#x2
417.
by jph00 4y ago
Yes exactly :D
418.
by jph00 4y ago
No it just uses normal git in the normal way. The simple trick is to use a jupyter-native git merge driver, so that merges are done at a cell level instead of a line level. Also, unneeded metadata is removed from the notebook when saving, s
419.
by jph00 4y ago
The pros and cons of Jupytext are discussed in the linked post. It's a great approach, but wasn't sufficient for our needs -- so for us, at least, it didn't fully solve the Jupyter+git problem. Specifically, it doesn't h
420.
by jph00 4y ago
What are the outstanding problems you feel are there even with the new nbdev2 functionality? Since I've been using it (the prerelease version) over the last few months I haven't come across a single problem, personally, despite do

More ›