Chief AI Scientist @cortAIx, Thales in Montreal - Canada. ex-@Accenture, ex-@IBM_Research, ex-@Orange. All views are my own!
Montreal, Canada
Joined January 2010
- Tweets1.3K
- Following130
- Followers529
- Likes449
Freddy Lecue retweeted
A very good paper I came across this morning by the @DeepMind researchers. For the past five years Transformers have been one of the most dominant approaches to Deep Learning problems, especially in the #NLP domain.
1/5
Freddy Lecue retweeted
After 2 years, Practical Deep Learning for Coders v5 is finally ready! 🎊
This is a from-scratch rewrite of our most popular course. It has a focus on interactive explorations, & covers @PyTorch, @huggingface, DeBERTa, ConvNeXt, @Gradio & other goodies 🧵
course.fast.ai/
Freddy Lecue retweeted
🧨Diffusion models have been powering impressive ML apps, enabling DALL-E or Imagen
Introducing 🤗 diffusers: a modular toolbox for diffusion techniques, with a focus on:
🚄Inference pipelines
⏰Schedulers
🏭Models
📃Training examples
github.com/huggingface/diffu…
Freddy Lecue retweeted
Congrats @h_nilforoshan, @jgaeb1, Ravi and @5harad on winning an outstanding paper award at ICML! #icml2022
Excited and honored to share that our #ICML2022 paper, “Causal Conceptions of Fairness and their Consequences” (w/ @jgaeb1, Ravi Shroff, and @5harad) has received an outstanding paper award!!! See 🧵for key findings and results
Paper: arxiv.org/abs/2207.05302 1/17
Freddy Lecue retweeted
🌸 BLOOM is here, available and accessible for everyone. 🤗🤝
Get yourself a huggingface.co account and try it yourself:
👉 huggingface.co/bigscience/bl…
At a glance :
🔢 176 billion parameter
🌍 59 languages
🔓 Open-Access
or read more about
👉bigscience.huggingface.co/bl…
ALT bloom widget
Freddy Lecue retweeted
A massive endeavour, hundreds of researchers, more than one year of hard work. From the community, for the community: BLOOM is here! 🌸
176B parameters, 59 languages, open access.
BLOOM is here. The largest open-access multilingual language model ever. Read more about it or get it at
bigscience.huggingface.co/bl…
hf.co/bigscience/bloom
Freddy Lecue retweeted
Happy to announce that there will be a workshop on Neuro Causal & Symbolic AI (nCSI) at @NeurIPSConf!
More information soon.
🖥️ ncsi.cause-lab.net/
Freddy Lecue retweeted
Official YOLOv7 (state-of-the-art real-time detector) is more accurate and faster than:
- YOLOv5 by 120% FPS
- YOLOX by 180% FPS
- Dual-Swin-T by 1200% FPS
- ConvNext by 550% FPS
- SWIN-L CM-RCNN by 500% FPS
- PPYOLOE-X by 150% FPS
arxiv.org/abs/2207.02696
github.com/WongKinYiu/yolov7…
Freddy Lecue retweeted
The OpenAI Minecraft paper is a great push to getting AI to work in Photoshop, Figma, or any software product — using just the keyboard & mouse, like a person would.
Steps in the paper, explained 🧵
1/
openai.com/blog/vpt/
Freddy Lecue retweeted
Wohoo, PyTorch 1.12 is out! Detailed release notes here: github.com/pytorch/pytorch/r….
Interesting (personal top 3) highlights
1) Updated DataPipes -- now with fully working shuffling and sharding (what are DataPipes? See my recent blogpost here: sebastianraschka.com/blog/20…)
[1/3]
Announcing PyTorch 1.12: TorchArrow, functorch, accelerated PyTorch training on Mac and more. Take a look: 👇
bit.ly/3OPnDbg
Freddy Lecue retweeted
Video2StyleGAN: Encoding Video in Latent Space for Manipulation
abs: arxiv.org/abs/2206.13078
Freddy Lecue retweeted
Statistics and mathematics are the backbones of data science and machine learning
Key statistical techniques 👇
Freddy Lecue retweeted
The meeting has been a few years in the making, but it was worth it! Thanks @lexfridman for the fascinating conversation - and the game of chess :) - it was a lot of fun!
Freddy Lecue retweeted
“Evolution through Large Models” – new paper from our team at OpenAI. Step towards evolutionary algos that continually invent and improve at inventing: Large models can suggest (+ improve at making) meaningful mutations to code. Paper: arxiv.org/abs/2206.08896 1/4
Freddy Lecue retweeted
📢 A 🧵on the future of NLP model inputs.
What are the options and where are we going? 🔭
1. Task-specific finetuning (FT)
2. Zero-shot prompting
3. Few-shot prompting
4. Chain of thought (CoT)
5. Parameter-efficient finetuning (PEFT)
6. Dialog
[1/]
Freddy Lecue retweeted
I'm excited to finally share what I have been working on.
Today we are officially launching Cohere For AI @forai_ml a non-profit research lab that aims to reimagine how, where, and by whom research is done.
cohere.for.ai/
Freddy Lecue retweeted
Robust-ResNet. 3 Key ideas to make CNNs more robust:
1) patchifying input images similar to ViTs;
2) increasing kernel size;
3) reducing activation layers & norm layers similar to ConvNext.
Paper: arxiv.org/abs/2206.03452
Code: github.com/UCSC-VLAA/RobustC…
