✨new blog post: Why don't academics use modern cloud services?
— Chris Holdgraf (@choldgraf) September 8, 2022
There were a lot of interesting ideas in this thread, so I wrote up a little blog post to summarize it! I hope it's useful:https://t.co/RoGuGZqlYe https://t.co/AUlRr7Qa76
On the Effectiveness of Compact Biomedical Transformers
— AK (@_akhaliq) September 8, 2022
abs: https://t.co/iwnSK5HSL7
huggingface models: https://t.co/vjdmrnfhgl
github: https://t.co/yerzCEukV8 pic.twitter.com/9h6c6RbdBq
Going down some deep rabbit holes here and learning new things ...
— Sebastian Raschka (@rasbt) September 7, 2022
Seems like a successful Kaggle strategy is randomly swapping cols in a tabular dataset (~like mix-up, but w/o including the labels).
Anyone tried this for a serious project with a non-deep learning tabular algo? pic.twitter.com/31amikDBNT
Part 2 of "Casual Forecasting at Lyft" is out! Excited they can share how we built a reliable, holistic model of Lyft's entire business.
— Sean J. Taylor (@seanjtaylor) September 6, 2022
Great post by my former teammates @DuaneJRich and @SameerSManek (+ @heng_kuang_ was heavily involved in the project).https://t.co/De6gCcCNBw
Stable Diffusion web UI with Outpainting, Inpainting, Prompt matrix, Upscale, Textual Inversion and many more features
— AK (@_akhaliq) September 5, 2022
github: https://t.co/F1T1qWMojs pic.twitter.com/hmjU4j1bic
George Hotz live-coding stable diffusion in tinygrad https://t.co/RR8Ewrdo5f via @YouTube
— Eric Jang (@ericjang11) September 5, 2022
Petals: Collaborative Inference and Fine-tuning of Large Models
— AK (@_akhaliq) September 5, 2022
abs: https://t.co/hE1Yx0P4iI
project page: https://t.co/Kgz6P4jZZx
github: https://t.co/dqcT4Ue3hh pic.twitter.com/5E1o2nTXsq
colab: https://t.co/VDRk57H3QU
— AK (@_akhaliq) September 4, 2022
“What is the explanation for why deep learning works? I understand backprop and gradient descent, but why should over-parametrized networks actually converge to anything useful?”https://t.co/uulgOtNBNz
— hardmaru (@hardmaru) September 4, 2022
More and more countries achieve to decouple economic growth from CO₂ emissions.
— Max Roser (@MaxCRoser) September 3, 2022
Just made this new chart that shows the evidence for 25 countries. Growth is up, emissions down. pic.twitter.com/8b1oAc44ad
What is the one resource I would recommend for anyone getting into RecSys?
— Radek Osmulski 🇺🇦 (@radekosmulski) September 2, 2022
This lecture by @xamat.
• it covers several foundational methods
• more importantly, it will teach you how to think about RecSys problems
Here are a couple of highlights:https://t.co/hCUWhQvWPs
COYO-700M, a large-scale dataset that contains 747M image-text pairs as well as many other meta-attributes to increase the usability to train various models
— AK (@_akhaliq) September 2, 2022
github: https://t.co/3J3hh5ocSu pic.twitter.com/nTbN9MztnD
