Radek Osmulski ๐Ÿ‡บ๐Ÿ‡ฆ (@radekosmulski) 's Twitter Profile
Radek Osmulski ๐Ÿ‡บ๐Ÿ‡ฆ

@radekosmulski

LLMs and retrieval by day and other genres of AI when I get the chance

๐Ÿงช Senior AI Eng @NVIDIAAI
๐Ÿซ @fastdotai trained DL Eng
๐Ÿ“ learn-dl.com

ID: 2435411576

linkhttps://radekosmulski.com/ calendar_today09-04-2014 12:54:38

8,8K Tweet

26,26K Followers

558 Following

Jia-Bin Huang (@jbhuang0604) 's Twitter Profile Photo

Why More Researchers Should be Content Creators Just trying something new! I recorded one of my recent talks, sharing what I learned from starting as a small content creator. youtu.be/0W_7tJtGcMI We all benefit when there are more content creators!

Why More Researchers Should be Content Creators

Just trying something new! I recorded one of my recent talks, sharing what I learned from starting as a small content creator. 

youtu.be/0W_7tJtGcMI

We all benefit when there are more content creators!
Dieter (@kagglingdieter) 's Twitter Profile Photo

Happy to finish 9th place solo gold in Kaggle wave inversion competition. For this competition I wrote my first custom CUDA kernel. My solution kaggle.com/competitions/wโ€ฆ

Hamel Husain (@hamelhusain) 's Twitter Profile Photo

I'm excited to announce Jeremy Howard & Jonathan Whitaker as guest speakers in our Evals course: bit.ly/evals-ai They'll showcase SolveIt, a VERY unique approach to software dev w/ AI. SolveIt borrows many ideas from notebooks: which is also our fav tool for evals ๐Ÿ˜

I'm excited to announce <a href="/jeremyphoward/">Jeremy Howard</a> &amp; <a href="/johnowhitaker/">Jonathan Whitaker</a> as guest speakers in our Evals course: bit.ly/evals-ai

They'll showcase SolveIt, a VERY unique approach to software dev w/ AI. SolveIt borrows many ideas from notebooks: which is also our fav tool for evals ๐Ÿ˜
JFPuget ๐Ÿ‡บ๐Ÿ‡ฆ๐Ÿ‡จ๐Ÿ‡ฆ๐Ÿ‡ฌ๐Ÿ‡ฑ (@jfpuget) 's Twitter Profile Photo

I don't do image competitions often but I did the seismic one on Kaggle. I ended up at 20th rank. Main miss from me was to not use additional training data as did top teams. Other than that it was a great learning experience. My writeup kaggle.com/competitions/wโ€ฆ

I don't do image competitions often but I did the seismic one on Kaggle. I ended up at 20th rank. Main miss from me was to not use additional training data as did top teams. Other than that it was a great learning experience. My writeup kaggle.com/competitions/wโ€ฆ
Jeremy Howard (@jeremyphoward) 's Twitter Profile Photo

Did you know we have a complete, heavily commented, updated idiomatic FastHTML app that shows how to create a login system, drag/drop, all CRUD endpoints, etc in a single small Python app? Read it yourself, or give it to your LLM as context. ๐Ÿ˜ github.com/AnswerDotAI/faโ€ฆ

Zach Mueller (@thezachmueller) 's Twitter Profile Photo

The weekend updates continue. On Tuesday I'm going to do a taxonomy on distributed training. DDP, PP, TP, DiLoCo, what on earth do these all mean? How do they work (at the high level), and how do you know which to use?

The weekend updates continue. 

On Tuesday I'm going to do a taxonomy on distributed training. DDP, PP, TP, DiLoCo, what on earth do these all mean? How do they work (at the high level), and how do you know which to use?
Zach Mueller (@thezachmueller) 's Twitter Profile Photo

With 14 world-expert speakers (and more on the way), free compute, and a great community, Scratch to Scale is the best course you can take right now when it comes to coming up to speed with how training at scale works today. Use the link below to get in at 35% off

With 14 world-expert speakers (and more on the way), free compute, and a great community, Scratch to Scale is the best course you can take right now when it comes to coming up to speed with how training at scale works today. Use the link below to get in at 35% off
Zach Mueller (@thezachmueller) 's Twitter Profile Photo

Pipeline Parallelism in a Nutshell Think of training a big model (multi-billion parameters) like a three-person kitchen: one friend prepares ingredients, one cooks, one plates the food. Instead of waiting for an entire dish to finish before starting the next, each friend

Jeremy Howard (@jeremyphoward) 's Twitter Profile Photo

We don't talk about Australian Masculinity enough. As a result, some young people have started relating to some very un-Australian un-manly points of view. Like the person below. Real men are responsible, independent, and care for their families, friends & communities.๐Ÿงต

We don't talk about Australian Masculinity enough.

As a result, some young people have started relating to some very un-Australian un-manly points of view. Like the person below.

Real men are responsible, independent, and care for their families, friends &amp; communities.๐Ÿงต
clem ๐Ÿค— (@clementdelangue) 's Twitter Profile Photo

Nvidia is killing it these days and sharing in the open for everyone's benefit! Currently second big organization on HF by number of repos created in the past 12 months (average a new repo every day) and crossed 30,000 followers (second in the US just after Meta).

Nvidia is killing it these days and sharing in the open for everyone's benefit!

Currently second big organization on HF by number of repos created in the past 12 months (average a new repo every day) and crossed 30,000 followers (second in the US just after Meta).