code_star's profile picture. Data Dawg @datologyai | Formerly Data Research Lead @DbrxMosaicAI | Visiting Researcher @ Facebook | Ph.D | #TXSTFOOTBALL fan | http://linktr.ee/code_star

Cody Blakeney

@code_star

Data Dawg @datologyai | Formerly Data Research Lead @DbrxMosaicAI | Visiting Researcher @ Facebook | Ph.D | #TXSTFOOTBALL fan | http://linktr.ee/code_star

置頂

We are looking for a post-training lead at @datologyai we have gpus, you can make them go brrrr

code_star's tweet image. We are looking for a post-training lead at @datologyai 

we have gpus, you can make them go brrrr

Its that time again

code_star's tweet image. Its that time again

Give it a year or two and we will have nanoagent. Then nanoagent speed run. Then nanoASI and so on.

Excited to release new repo: nanochat! (it's among the most unhinged I've written). Unlike my earlier similar repo nanoGPT which only covered pretraining, nanochat is a minimal, from scratch, full-stack training/inference pipeline of a simple ChatGPT clone in a single,…

karpathy's tweet image. Excited to release new repo: nanochat!
(it's among the most unhinged I've written).

Unlike my earlier similar repo nanoGPT which only covered pretraining, nanochat is a minimal, from scratch, full-stack training/inference pipeline of a simple ChatGPT clone in a single,…


Cody Blakeney 已轉發

Here’s a scaling law that matters more: every two years the gap to frontier closes by a factor of 2x

Excited to release new repo: nanochat! (it's among the most unhinged I've written). Unlike my earlier similar repo nanoGPT which only covered pretraining, nanochat is a minimal, from scratch, full-stack training/inference pipeline of a simple ChatGPT clone in a single,…

karpathy's tweet image. Excited to release new repo: nanochat!
(it's among the most unhinged I've written).

Unlike my earlier similar repo nanoGPT which only covered pretraining, nanochat is a minimal, from scratch, full-stack training/inference pipeline of a simple ChatGPT clone in a single,…


I’m fairly convinced the first rabbit in a hat trick was an accident


code_star's tweet image.
code_star's tweet image.

New website looking pretty slick 👀

code_star's tweet image. New website looking pretty slick 👀


code_star's tweet image.

Talk from Wenting Zhao of Qwen on their plans during COLM. Seems like 1 word is the plan still: scaling training up! Let’s go.

natolambert's tweet image. Talk from Wenting Zhao of Qwen on their plans during COLM. Seems like 1 word is the plan still: scaling training up! Let’s go.


Doing some on the ground research on power (football) analysis

Was watching the Georgia game and noticed @dwarkesh_sp, Host of @dwarkeshpodcast was literally on TV 🤯🤯🤯🤯🤯



Cody Blakeney 已轉發

One day, @code_star will swap profile pics with @willccbb and the world will tremble.

One thing I wonder is, these GPUs are going to remain incredibly powerful and the next generation or two of GPUs are going to be rolled at scales we have never really seen before. Now obviously electricity isn't free, but if we find ourselves with much cheaper electricity in…



Quit hoggin'em all

Each dot represents 5,000 hogs

uncledoomer's tweet image. Each dot represents 5,000 hogs


One thing I wonder is, these GPUs are going to remain incredibly powerful and the next generation or two of GPUs are going to be rolled at scales we have never really seen before. Now obviously electricity isn't free, but if we find ourselves with much cheaper electricity in…

So in like 5 ish years ... what is going to happen to all the H100s ... landfill?



what could this possibly mean ...

code_star's tweet image. what could this possibly mean ...

Has something changed … for the better on the algorithm? Definitely noticing more engagement and longer lived tweets. Is nature healing?


Cody Blakeney 已轉發
allgarbled's tweet image.

just sf tech party things: 3 floors of djs Sam Altman pics on the wall Coworking room (fully packed) Merch trucks (Not pictured) a girl dressed up as an astronaut Props to @BellaNazzari for the very walm welcome to the city 🥂

SofiGu1's tweet image. just sf tech party things:

3 floors of djs
Sam Altman pics on the wall
Coworking room (fully packed)
Merch trucks
(Not pictured) a girl dressed up as an astronaut

Props to @BellaNazzari for the very walm welcome to the city 🥂
SofiGu1's tweet image. just sf tech party things:

3 floors of djs
Sam Altman pics on the wall
Coworking room (fully packed)
Merch trucks
(Not pictured) a girl dressed up as an astronaut

Props to @BellaNazzari for the very walm welcome to the city 🥂
SofiGu1's tweet image. just sf tech party things:

3 floors of djs
Sam Altman pics on the wall
Coworking room (fully packed)
Merch trucks
(Not pictured) a girl dressed up as an astronaut

Props to @BellaNazzari for the very walm welcome to the city 🥂
SofiGu1's tweet image. just sf tech party things:

3 floors of djs
Sam Altman pics on the wall
Coworking room (fully packed)
Merch trucks
(Not pictured) a girl dressed up as an astronaut

Props to @BellaNazzari for the very walm welcome to the city 🥂


Lies

I don’t care what fashion designers say, grown men cannot actually wear corduroy pants embroidered with little dogs to work.

castlehillmom's tweet image. I don’t care what fashion designers say, grown men cannot actually wear corduroy pants embroidered with little dogs to work.


Cody Blakeney 已轉發

<smol> is the way to go

Does @thinkymachines call them <thinky> tokens?



So in like 5 ish years ... what is going to happen to all the H100s ... landfill?


Cody Blakeney 已轉發

going viral on link*din feels unnatural! but you know what, OpenMed loves to be everywhere! 😈

MaziyarPanahi's tweet image. going viral on link*din feels unnatural!

but you know what, OpenMed loves to be everywhere! 😈

It’s a great day to watch some football and some loss curves go down


code_star's tweet image.
code_star's tweet image.

If gradstudents knew what actually worked in training SOTA LLMs they would be so mad



Loading...

Something went wrong.


Something went wrong.