Interesting demos:
- https://huggingface.co/spaces/Nexusflow/NexusRaven-V2-Demo
- https://huggingface.co/spaces/yunyangx/EfficientSAM
- https://huggingface.co/spaces/openai/whisper
- https://huggingface.co/spaces/diffusers/unofficial-SDXL-Turbo-i2i-t2i
- https://huggingface.co/blog/personal-copilot
Set up: Google Colab: https://colab.research.google.com/ - 16GB T4 GPU.
- We start by covering the fundamentals of Prompting. We utilize the
mistralai/Mistral-7B-Instruct-v0.2
open access model, which excels at following instructions. We demonstrate how to enable token streaming and employ chat templates. - We dive into powerful prompting techniques like few-shot, chain-of-thought (CoT), Self Consistency, ReACT (Reason & Act), and Tree of Thoughts (ToT) with coding examples. We'll showcase a Breadth First Search based ToT example and Langchain tools for ReACT example.
- Explore JSON-only outputs, prompt chaining, and LLMs for evaluation. Plus, discover how to create a Gradio chatbot 🚀.
- We will see how Langchain makes it easy as a framework to bring together most of the steps aboves in simple APIs. We will be using open source models from 🤗 Hub, Langchain and Chroma vector DB.
POC: https://huggingface.co/spaces/smangrul/PEFT-Docs-QA-Chatbot