Nemotron, every mention
29 scenes, the whole family · ← back to Nemotron
tap a year for its mentions
every year anyone Matt Turck 36Bryan Catanzaro 35Dan Fu 1
Verbatim, from the transcripts: the passages where Nemotron comes up
“OpenAI’s Model Hacked Us” - Hugging Face’s Thomas Wolf
- ▶ 43:41 Matt Turck But if, if you think of the West, then, so you mentioned Nvidia and we had, uh, Brian Catanzaro, um, uh, from, you know, the whole NemoTron effort on the podcast uh, a few weeks ago.
Inside Nemotron & NVIDIA’s AI Lab | Bryan Catanzaro
- ▶ 0:43 Matt Turck Brian Catanzaro leads NemoTron, NVIDIA's family of open foundation models. 2 times in the scene
- ▶ 0:56 Matt Turck And Nemotron III Ultra immediately became the number one US open weights model when it was released just a couple of weeks ago.
- ▶ 1:39 Matt Turck So you guys at NVIDIA just released Nemo
- ▶ 10:10 Bryan Catanzaro Um, and you know, we're pushing Nemo Tron along here at Nvidia as well.
- ▶ 12:40 Matt Turck I'd love to go into a bit of a deep dive into NemoTron, but before we do that, maybe a few minutes on your story, your background.
- ▶ 21:43 Bryan Catanzaro Um, and also led to the foundations of today's NemoTron project, um, where, you know, Nvidia trains, uh, its own LLMs, um, uh, for its own purposes.
- ▶ 21:56 Matt Turck So let's go into all things, uh, NemoTron. 8 times in the scene
- ▶ 26:58 Bryan Catanzaro You know, the original, what, what, uh, we originally called Mnemotron One was actually a project that we did with Microsoft. 2 times in the scene
- ▶ 27:34 Bryan Catanzaro Uh, we got up to NemoTron three. 3 times in the scene
- ▶ 28:27 Bryan Catanzaro We released a, a NemoTron two, um, I believe it was last year. 2 times in the scene
- ▶ 28:57 Bryan Catanzaro Um, now we're in a, a slightly difficult state because, you know, we're working on Nemotron IV, right? 4 times in the scene
- ▶ 30:33 Bryan Catanzaro You know, we followed through over 10 plus years with CUDA, and we're doing that with Nemo Tron now. 8 times in the scene
- ▶ 33:38 Matt Turck What's the current state of the Neumontron family? 6 times in the scene
- ▶ 33:42 Matt Turck You got Nano, you got Super, you got Ultra. 3 times in the scene
- ▶ 33:42 Matt Turck You got Nano, you got Super, you got Ultra. 4 times in the scene
- ▶ 35:36 Bryan Catanzaro NemoTron, ah, three family has a lot of things in it that are, ah, we're really proud of.
- ▶ 39:27 Matt Turck So as we, uh, get into a slightly more technical things, the, uh, architecture of Mnemotron, uh, is hybrid. 2 times in the scene
- ▶ 45:37 Bryan Catanzaro This is speaking to Nematron's first job.
- ▶ 46:00 Bryan Catanzaro Latent MOE is a specific, uh, innovation that we have in NemoTron three family.
- ▶ 47:26 Matt Turck Another important characteristic of Numatron, free ultra is a one million token context, the, the long context window. 2 times in the scene
- ▶ 51:48 Bryan Catanzaro Um, so with, you know, um, uh, with, uh, our recent Neotron models, you know, we're pretty proud of our acceptance rates, but we're always trying to make them better.
- ▶ 52:54 Matt Turck Uh, what does that mean in the context of, uh, Nemo Tron three? 2 times in the scene
- ▶ 52:59 Bryan Catanzaro So with Nemo Tron three ultra, we did post-training using something called multi-domain on policy distillation.
- ▶ 54:58 Bryan Catanzaro And so this particular technology, um, has been really instrumental in helping more people work together to make Nemotron stronger.
- ▶ 55:24 Matt Turck One of the, uh, exciting things that you all did, uh, in the context of NemoTron is also to publish the data, the training data. 3 times in the scene
- ▶ 1:00:53 Bryan Catanzaro And my team is not the only team building NemoTron. 7 times in the scene
- ▶ 1:04:43 Bryan Catanzaro Um, uh, inside NemoTron, uh, you know, so we, we have a budget, uh, for NemoTron, um, and inside NemoTron, we allocate compute based on what we think the needs of the project are.