Mougatine

6 ms·

1.
Fault Tolerant Llama training(pytorch.org)
66 points by Mougatine 1y ago|14 comments
2.
MuLoCo: Muon is a practical inner optimizer for DiLoCo(arxiv.org)
2 points by Mougatine 1y ago|0 comments
3.
by Mougatine 1y ago
very cool work!
4.
OpenDiLoCo: Open-Source Framework for Distributed Low-Communication Training(arxiv.org)
4 points by Mougatine 2y ago|0 comments
5.
by Mougatine 3y ago
https://arthurdouillard.com/deepcourse/
6.
by Mougatine 3y ago
"From scratch" is commonly used in the field.
7.
by Mougatine 4y ago
In a way, it's similar to the actors selling their digital images in the movie The Congress. It's inevitable.
8.
by Mougatine 4y ago
You don't need a PhD for DS/ML Engineer. Even at FAANG, even in their research labs. Usually a PhD is only in the requirements for Research Scientist (RS). That said, I did a PhD (and am now a RS). It's a fantastic opportunit
9.
by Mougatine 4y ago
You could really raise money (millions?) in less than 3 days before? That sounds crazy to me.
10.
by Mougatine 4y ago
https://arthurdouillard.com/deepcourse/ It's a deep learning course I built for my students around some kinds of "skill tree" like in video games. PS: I know CSS can be a bit broken on some browsers, I d
11.
by Mougatine 5y ago
That sounds nice! I'll try it. Is it possible to then export the cards to Anki?
12.
by Mougatine 5y ago
A $100,500 bounty seems pretty cheap compared to the severity of the issue, or is it common?
13.
by Mougatine 5y ago
Thanks, for the remarks, I'll fix that.
14.
by Mougatine 5y ago
Shameless plug: I've made a deep learning course oriented with practical content on a wide variety of computer vision topics --> https://arthurdouillard.com/deepcourse/ with slides, google colab, and anki cards
15.
Show HN: Deep Learning for Computer Vision course with colabs and Anki cards(arthurdouillard.com)
12 points by Mougatine 5y ago|2 comments
16.
by Mougatine 5y ago
I've built with a collegue Continuum: https://github.com/Continvvm/continuum It's a Python library giving utilities to handle data in Continual Learning, a subfield of Machine/Deep-learning where data co
17.
by Mougatine 6y ago
But they weren't as good. e.g. look at translation. It was not usable a decade ago
18.
Continual Learning at CVPR 2020(medium.com)
1 points by Mougatine 6y ago|0 comments
19.
Operation Red Falcon (2015)(magazine.atavist.com)
1 points by Mougatine 6y ago|0 comments
20.
Lifelong Learning for Deep Neural Networks (2019)(medium.com)
2 points by Mougatine 7y ago|0 comments
21.
by Mougatine 7y ago
Probably because Algolia isn't using yet any machine learning, but more old-school (yet still very good) tries-based search. And it's probably why also Algolia shines as search engine for a website vs searching the internet like G
22.
by Mougatine 7y ago
I'm French (from France) and it's also perfectly fine to say that. I've never heard this sentence used to say anything else...
23.
Seeing Is Not Necessarily Believing: Limitations of GANs for Data Augmentation(openreview.net)
2 points by Mougatine 7y ago|0 comments
24.
Cars detection from satellite imagery with RetinaNet(medium.com)
2 points by Mougatine 8y ago|0 comments
25.
Human or Company(twitter.com)
1 points by Mougatine 8y ago|0 comments
26.
3 Small but Powerful Convolutional Networks(towardsdatascience.com)
4 points by Mougatine 8y ago|0 comments
27.
An Explanation of Densely Connected Convolutional Networks(arthurdouillard.com)
1 points by Mougatine 8y ago|0 comments
28.
Amazon launches an Android app in India called “Internet”(techcrunch.com)
1 points by Mougatine 8y ago|0 comments
29.
Summary of “Deep Learning Scaling Is Predictable, Empirically”(arthurdouillard.com)
1 points by Mougatine 8y ago|0 comments
30.
How RoI Pooling and RPN Work in Faster-RCNN(arthurdouillard.com)
2 points by Mougatine 8y ago|0 comments

More ›