←
Jack Cook168 posts
Jack Cook avatar✓

Jack Cook ✓

@jackcookjack

mts @thinkymachines | ex-phd student @mit | systems for ml

◷ Joined January 2011▧ 42 Photos and videos
Jack Cook avatar
Jack Cook@jackcookjack·Nov 13

Inkling 2 Small is our first model trained partly in NVFP4, with a mixed-precision recipe that enables a great tradeoff between speed and accuracy. Check out our new post to see how we did it, and the lessons we learned along the way!

Mira Murati avatar
⟳ Jack Cook retweeted
Mira Murati@miramurati·Nov 13

Today on Connectionism: how we trained Inkling 2 Small with a native 4-bit forward pass on @nvidia GB300s, from the first pretraining step through the final RL rollout, one of the many improvements across the stack that took pretraining from weeks to 68 hours.

NVIDIA AI avatar
⟳ Jack Cook retweeted
NVIDIA AI@NVIDIAAI·Nov 13

Congrats @thinkymachines team on Inkling 2 Small! A 594B model pretrained in just 68 hours with NVIDIA NVFP4 on GB300 NVL72 racks. Read how the team made low-precision training work at this scale. 👇

Jakeup avatar
⟳ Jack Cook retweeted
Jakeup@myhandle·Nov 12

ML researcher: ...then we must de-quantize the tensors. Me: That's not a real thing, that's some gibberish the villain says in a cheap sci-fi movie. ML researcher: ... Claude: That just means storing each number with more memory bits. Me: Oh okay, why didn't you say so.

Han Guo avatar
⟳ Jack Cook retweeted
Han Guo@HanGuo97·Nov 13

Whenever I see the bigram Jack + NVFP4, I know it has to be a cool work. Congrats on the release!

Paul Zhang avatar
⟳ Jack Cook retweeted
Paul Zhang@pz_ai1·Nov 13

@jackcookjack is a low precision legend. Give this a read to for a sense of how we do nvfp4 pretraining and RL at thinking machines!

Guangxuan Xiao avatar
⟳ Jack Cook retweeted
Guangxuan Xiao@Guangxuan_Xiao·Nov 13

the NVFP4 recipe behind Inkling 2 Small: 4-bit forward from the first pretraining step to the final RL rollout, MXFP8 + stochastic rounding for backward and still more to come 🪜

Horace He avatar
⟳ Jack Cook retweeted
Horace He@cHHillee·Nov 13

When I first talked to @jackcookjack, he thought that although he was having fun working on nvfp4 pretraining in academia, he was skeptical that it would ever actually work for actual large-scale runs. As it turns out, he was wrong :^) This blog post explains how it can be done

Soumith Chintala avatar
⟳ Jack Cook retweeted
Soumith Chintala@soumithchintala·Nov 5

Our v2 small model is here. I've been using it daily for a variety of things. The pre-training ran for 68 hours on GB300s. Post-training was 17 million rollouts over 9 days. Entirely pre-mid-post trained in NVFP4. Available to train and fine-tune on Tinker right now. Open weights next week after external safety testing is fully complete (cant skimp on that). Continuing to decentralize and democratize intelligence...

Thinking Machines avatar
⟳ Jack Cook retweeted
Thinking Machines@thinkymachines·Nov 5

Introducing Inkling 2 Small. Scoring 31 on the AA Intelligence Index, it's a big jump over its predecessor. Inkling 2 Small is available for fine-tuning and inference on Tinker today, with full weight release planned soon. thinkingmachines.ai/news/ink…

Thinking Machines avatar
⟳ Jack Cook retweeted
Thinking Machines@thinkymachines·Jul 15

Today, we are introducing Inkling. Inkling reasons efficiently across text, image, and audio modalities. We are making the full weights available. thinkingmachines.ai/news/int… Available today for fine-tuning on Tinker. Play with it in the Inkling Playground. 🧵