Phi-2: Small Language Model Performance Tutorial - Google Colab with Python

Explore the capabilities of Microsoft Research's Phi-2, a 2.7 billion-parameter language model that outperforms larger models, in this comprehensive tutorial video. Learn about the innovative training methods behind Phi-2, including 'textbook-quality' data and scaled knowledge transfer techniques. Follow along as the model is loaded into Google Colab and tested across various tasks such as coding, math, reasoning, and data extraction. Gain insights into Phi-2's performance compared to Mistral and Llama 2, and understand its potential impact on the field of language models. Discover how to set up the environment, use the correct prompt format, and generate text effectively. Dive into practical examples of math problem-solving, coding tasks, and text analysis using Phi-2.

- Intro
- AI Bootcamp on MLExpert.io
- What is Phi-2?
- Phi-2 vs Mistral vs Llama 2
- Phi-2 on HuggingFace Hub
- Google Colab Setup
- Prompt Format
- Text Generation
- Math
- Coding
- Text Analysis
- Conclusion