Week 1

Finished

• Studied CWRU HPC Orientation, including:
  • graphical access
  • work nodes requesting
  • job scheduling and submitting
  • etc.
• Got familiar with FrameNet dataset, including:
  • FrameNet online dataset
  • local dataset stored on HPC
    • /mnt/rds/redhen/gallina/projects/ChattyAI/FramesConstructions/fndata-1.7
• Learned about available open-source LLM models, including:
  • LlamaIndex
  • Llama 2
• Started thinking about fine-tuning methods and RAG
  • Crafting Your Own Dataset for Fine-Tuning Llama2 in Google Colab: A Step-by-Step Guide
  • Fine-Tuning LLaMA 2: A Step-by-Step Guide to Customizing the Large Language Model
  • Step by step no-code RAG application using Langflow.
  • Building Agentic RAG with LlamaIndex

Other Messages From The Weekly Meeting

• Before working onto something bigger, finish something smaller but works. Make every step steady so as to build a successful project.
• FrameNet is limited (e.g. no frames examples for tourism),

Challenges

• What would be the proportion of fine-tuning and RAG?
• How to build the evaluation matrix / how to assess the accuracy of the generated result?
  • How to build the system so that it doesn’t just randomly throw 2 frames together in a terrible way that everyone would think it is a disaster?

---

<

Previous Post

Project Details

>

Next Post

Week 2