How Llama 3.1 Works
Автор: Oxen
Загружено: 2024-08-26
Просмотров: 734
Описание:
In this dive we go into the absolute behemoth of a paper (92-pages) Llama 3 Herd of Models. We look at how Meta created the most competitive open-source model to date.
--
Get Oxen AI 🐂 https://oxen.ai/
Oxen AI makes versioning your datasets as easy as versioning your code! Even is millions of unstructured images, the tool quickly handles any type of data so you can build cutting-edge AI.
--
Paper 📜 https://ai.meta.com/research/publicat...
Links + Notes 📝 https://www.oxen.ai/blog/arxiv-dives
Join Arxiv Dives 🤿 https://oxen.ai/community
Discord 🗿 / discord
--
Chapters
0:00 Intro
5:39 The Three Levers to Improve Llama 2
9:05 Pre-Training vs Post-Training
29:24 Post Training
31:50 Reward Model and Language Model
36:17 SFT Data
17:10 Synthetic Data Quality
41:30 Capabilities
43:10 Code
Повторяем попытку...
Доступные форматы для скачивания:
Скачать видео
-
Информация по загрузке:
