Skip to main content
EleutherAI

EleutherAI

Search complete. 22 mentions across 3 episodes found for "EleutherAI".

Nov 9, 2026

Siddharth KaramchetiGUEST
61:10
This happened every time you're training any big model in the 1 billion plus parameter range.
Siddharth KaramchetiGUEST
61:16
So the Eleuther AI folks, this happens all of the time.
Siddharth KaramchetiGUEST
61:19
The T5 models have this thing in one of the notes in the GitHub repository for training a T5, which is, if training fails, rewind to the latest checkpoint, re-randomize your data order, and then try again, and it won't crash.
Siddharth KaramchetiGUEST
61:34
And that's kind of how most modern language models are trained right now.
Harrison KinsleyHOST
7:28
If I forget and you're looking for it and you can't find it, just leave a comment and I'll link to that.
Harrison KinsleyHOST
7:32
So for our very first instruct model, let's check out the smaller, faster, and super cheap at... 0.0002, hopefully I got that right, per 1000 tokens model from Together, which was fine-tuned from the GPT-JT model, which is trained from the GPT-J model from Eleuther AI.
Harrison KinsleyHOST
7:52
It's kind of crazy how even AI models themselves are now layering on top of each other and abstracting the models themselves.
Harrison KinsleyHOST
7:59
So the model path for this model is togethercomputer slash redpajama dash insight7binstruct.

1 more episode mentions EleutherAI.

Create an account to see the whole feed, search across every transcript, and follow the entities you care about.

We value your privacy

We use cookies to understand how you use our platform and to improve your experience. Click “Accept All” to consent, or “Decline non-essential” to opt out of non-essential cookies. Read our Privacy Policy.