Published On Jun 24, 2024
#aiagents #moa #llm #mistral
This video is an easy explanation of the Mixture of Agents (MoA) method and algorithm and a tutorial on how to run a MoA multi-LLM AI agents locally and 100% FREE. The method is discussed in this paper:
Wang et al. (2024). Mixture-of-Agents Enhances Large Language Model Capabilities.
Full process to run the MoA system in this GitHub repo:
https://github.com/Maryam-Nasseri/MoA...
Tutorial to set up an agentic workflow with CrewAI and Ollama:
• 💯 FREE Local LLM - AI Agents With Cre...
Explanation of various agentic systems & AI agents:
• 🔴 This Agentic AI Workflow Will Take ...
How Hugging Face evaluates LLMs:
• What Language Model To Choose For You...
Chapters and Key Terms:
00:00 Introduction to Mixture of Agents (MoA)
00:32 MoA architecture and layers of agents
02:07 Automatic model selection: Performance & Diversity
03:10 Proposer and Aggregator agents
04:00 MoA evaluation with LLAMA3, Mixtral, and Qwen1.5 against GPT- 4/GPT-4o
04:18 Benchmarks: AlpacaEval 2.0, MT-bench, FLASK
05:14 Together AI API key set-up as VENV variable
05:42 Clone git and install dependencies
06:10 MoA algorithm's main components
06:59 Run MoA with Qwen2, Qwen 1.5, Mixtral, and dbrx from Databricks
07:33 Solving a problem from the GSM8K benchmark used in the Hugging Face Leaderboard
/ @analyticscamp