Kone ismael retweeted
Quick demo of the voice to voice translation between English and Twi that you can try today on our web app.
This one is from the perspective of a tourist visiting Ghana and needing to get around.
Brought to you by the amazing team at @GhanaNLP @algorineinc
Kone ismael retweeted
I had to travel 26 hours and spend $2000+ to join #ICLR2023 in Rwanda.
But people in Africa have to do this every time a conference is held in US.
What happens when we make it easier to participate?
1530% higher registrations from Africa.
This is important and must continue.
Kone ismael retweeted
We’ve trained language models to be better at responding to adversarial questions, without becoming obtuse and saying very little. We do this by conditioning them with a simple set of behavioral principles via a technique called Constitutional AI: anthropic.com/constitutional…
Kone ismael retweeted
Within a day, timm will be past the 20k mark in GitHub stars. It's been quite the journey getting here. Thanks to support of community, incl users and contributers that helped out with models, bug fixes, reports, helping other users. github.com/rwightman/pytorch…
Kone ismael retweeted
New #fastai release: v2.7.5. github.com/fastai/fastai/rel…
New Features
- auto-normalize timm models
- PyTorch 1.12 support
Kone ismael retweeted
Announcing decoder denoising pretraining for semantic segmentation: arxiv.org/abs/2205.11423
Take a U-Net, pretrain the encoder on classification, pretrain the decoder on denoising, and fine-tune on semantic segmentation. This achieves a new SoTA on label efficient segmentation.
Kone ismael retweeted
NVIDIA Deep Learning Institute has launched a new free self-paced course, Disaster Risk Monitoring Using Satellite Imagery. Learn how to build and deploy a deep learning model to automate the detection of flood events using satellite imagery. Enroll now! bit.ly/3u3S8m9
Kone ismael retweeted
2022 will mark the return for the @fastdotai course of neural nets implemented in spreadsheets!
Kone ismael retweeted
I am over the moon to announce:
1) I'm now a professor at University of Queensland (UQ), the top institute in my home state!
2) I'll be teaching a brand new deep learning course at UQ from April, which will form the basis of a new @fastdotai course! 🧵
eait.uq.edu.au/article/2022/…
Kone ismael retweeted
The @fastdotai community is hosting a documentation sprint!
This is a great opportunity to contribute to an open-source library!
It takes place Feb 14th-26th!
@TheZachMueller will host a kickoff livestream at 3pm PST & more details will be shared soon
youtu.be/GwDVlR-eZwc
Kone ismael retweeted
For my midterm I read almost all of the Deep Learning book, until my finals (in two months) I'm writing a very comprehensive blog post that contains my notes and the TF Keras code for each architecture, how (ELI5 maths) and why they work, where to use which 🤓
Kone ismael retweeted
I want to retract this tweet publicly:
xitter.cf/_clashluke/status/1463…
It is incorrect and causes unnecessary harm to the authors of "PoolFormer: MetaFormer is Actually What You Need for Vision" (arxiv.org/abs/2111.11418).
Using just AvgPool and MLP, they outperform most models.
1/6
Kone ismael retweeted
What a cool project!
Plus the code is available and there is a step by step walkthrough in the Jupyter NB 🤩
Not sure how it is possible, but the blog post is even better 😁
Model training is done with @fastdotai!
What a treasure! Added to my reading list.
I used Deep Learning to create a Real Time Sign Language Classifier that runs on the webcam. The whole project is done from scratch.
You can find all the code in my github repo: github.com/jimmiemunyi/Sign-…
Here is a blog post on the same topic:
: jimmiemunyi.github.io/blog/ 😃😃
Kone ismael retweeted
Updated version of the Intermediate Subscription is live! Video workbooks with vocab and interactive exercises for anyone ready to go beyond the basics in Bambara/Jula :-)
ankataa.com/intermediate-sub…
Kone ismael retweeted
The new chapter in the Python array interoperability story.
We (me and my colleagues at @quansightai together with the library maintainers) are planning to add GPU & distributed arrays support for SciPy, scikit-learn, and scikit-image.
Find out more: labs.quansight.org/blog/2021…
Kone ismael retweeted
Happy to announce that our paper, "The Low-Resource Double Bind: An Empirical Study of Pruning for Low-Resource Machine Translation," was accepted to Findings of EMNLP 2021! Joint work with @sarahookr and @KreutzerJulia
📜openreview.net/forum?id=qw1q…
1/n
Kone ismael retweeted
I endorse this message :)
We encourage you to register for the GTC conference, which will take place on November 8-11, 2021. you get to learn about the absolute latest AI on the planet & get inspired.
It’s a unique experience that you wouldn’t want to miss out on.
nvidia.com/gtc/sessions/emer…
Kone ismael retweeted
We all know NNs are super expressive, so why is it that models seem to make similar mistakes?
In our new work, we investigate how training methodology⚙️ affects learned representations🧠 and prediction behavior🏃🏽♀️
📄 arxiv.org/pdf/2110.12899.pdf
🧵 Read on for the surprises we found!
