Explore the key changes and improvements in LLaMA2, Meta's latest large language model. Dive into the three base models, examine the LLaMA paper, and compare win rates against other models. Learn about the release of 7B, 13B, and 70B models, and understand the decision to delay the 34B model. Analyze pre-training graphs for different models and discover the role of third-party labeling and annotation. Compare RLHF implementations across various models and get hands-on with the Hugging Face LLaMA 70B chatbot. Access resources including Meta's website, HuggingFace models, and GitHub repositories for further exploration.
Overview
Syllabus
Intro
Three LLaMA Base Models
LLaMA Paper
Win Rate of LLaMA vs. Other Models
Released of Models 7B, 13B, 70B and Delaying 34B Model
Graph of Pre-training Different Models
Third Party to do Labeling and Annotating
Comparing what other people have done with RLHF
Hugging Face LLaMA 70B Chatbot
Taught by
Sam Witteveen
Related Courses
-
Investigating the Code Llama Models and Fine-Tuned Alternatives
-
LLaMA2 Tokenizer and Prompt Tricks
-
Llama 2 in LangChain - First Open Source Conversational Agent
-
Llama 3 - 8B and 70B Models Deep Dive and Implementation Guide
-
Open LLaMa in LangChain and Hugging Face
-
Better Llama with Retrieval Augmented Generation - RAG
4.0
