Notebooks & Instructions

Getting Started

Note: most of the examples use the OpenAI API or Llama.  These decisions were made out of convenience (to minimize the number to distinct setups), and students are allowed to use any LLM API they find most beneificial.

Setting up OpenAI API:

Our Colab Notebooks will use GPT API calls which means student will need to get an API Key. Follow this website tutorial https://www.maisieai.com/help/how-to-get-an-openai-api-key-for-chatgpt under the "How do I get an OpenAI API Key?" section (noticing that the menu of your OpenAI account, where you can find [API keys] and [Billing] (located under [Settings]), is currently on the left sidebar on the latest OpenAI website layout). 

Remember we will reimburse $50 per student for API usage.

Choosing a model:

Models vary in terms of their number of parameters, what datasets they were trained on, their context lenght, and how they have been tuned (e.g., instruction tuning).  You can see some of the considerations here.  Note that larger models are more expensive (GPT4 is generally much more expensive that GPT 3.5).  The same type of analysis applies if you wish to use Gemini or Claude.

Fine-Tuning:

Many companies offer API functionality for fine-tuning models.  

• OpenAI GPT 

• Google Gemini 

Ollama:

You can also host your own open source models locally.  Ollama is a popular tool for this setting.

Colab Notebooks

Notebook 1 -- Simple example to get started.

Simple Agent Notebook -- Pandas manipulations