generative-ai-for-beginners/18-fine-tuning/SETUP.md

Setup Your Dev Environment

We have instrumented this repository with a dev container that comes with a Python 3 runtime. Simply open the repo in GitHub Codespaces or on your local Docker Desktop, to activate the runtime automatically. Then open th Jupyter notebook and select the Python 3.x kernel to prepare the Notebook for execution.

1. Create .env file

The default notebook is set up for use with an Azure OpenAI service resource. To configure this, we need to setup local environment variables for Azure as follows:

1. Look in the root folder for a .env.copy file. It should contain a list of name-value pairs like this:

   AZURE_OPENAI_ENDPOINT='<add your endpoint here>'
   AZURE_OPENAI_DEPLOYMENT='<add your deployment name here>'
   AZURE_OPENAI_KEY='<add your key here>'
   AZURE_OPENAI_EMBEDDINGS_DEPLOYMENT='<add your deployment name here>'
   

2. Make a copy of that file called .env using a command like this at the terminal:

   cp .env.copy .env
   

   This should create an identical copy except that this file is .gitignore-d and will never get checked into source control. We can now populate this .env file with the environment variable values (secrets) without fear of them being checked in accidentally. You can now move to the next section to start populating these variables.

3. (Option) If you use GitHub Codespaces, you can also save environment variables as Codespaces secrets as described here. Visit that link to learn how to add a secret, edit it, delete it, and use it in your running codespace. If you choose this option, your GitHub Settings > Codespaces Secrets should look something like this.

   Note that you can create these secrets once and then make them available to multiple repositories that use those env variables. However, note that this option works only if you use GitHub Codespaces. You will still need to setup the .env file if you plan to use Docker Desktop for local development with that devcontainer configruation.

2. Populate .env file

Let's take a quick look at the variable names to understand what they represent:

Variable	Description
AZURE_OPENAI_ENDPOINT	This is the deployed endpoint for an Azure OpenAI resource
AZURE_OPENAI_KEY	This is the authorization key for using that service
AZURE_OPENAI_DEPLOYMENT	This is the text generation model deployment endpoint
AZURE_OPENAI_EMBEDDINGS_DEPLOYMENT	This is the text embeddings model deployment endpoint

For context, the last two variables refer to specific models that are used in chat completion (text generation model) and vector search (embeddings model) activities that are frequently used in generative AI applications. In the following sections, we'll locate the values for these variables and set them in .env (replacing the content within the ' ', but preserving the quotes).

2.1 Use Azure Portal

The Azure OpenAI endpoint and key values will be found in the Azure Portal so let's start there.

1. Navigate to the Azure Portal
2. Click the Keys and Endpoint option in the sidebar (menu at left).
3. Click Show Keys - you should see the following: KEY 1, KEY 2 and Endpoint.
4. Use the KEY 1 value for AZURE_OPENAI_KEY
5. Use the Endpoint value for AZURE_OPENAI_ENDPOINT

Next, we need the endpoints for the specific models we've deployed.

1. Click the Model deployments option in the sidebar (left menu) for Azure OpenAI resource.
2. In the destination page, click Manage Deployments

This will take you to the Azure OpenAI Studio website, where we'll find the other values as described below.

2.2 Use Azure OpenAI Studio

1. Navigate to Azure OpenAI Studio from your resource as described above.
2. Click the Deployments tab (sidebar, left) to view currently deployed models.
3. If your desired model is not deployed, use Create new deployment to deploy it.
4. You will need a text-generation model - we recommend: gpt-35-turbo
5. You will need a text-embedding model - we recommend text-embedding-ada-002

Now update the environment variables to reflect the Deployment name used. This will typically be the same as the model name unless you changed it explicitly. So, as an example, you might have:

AZURE_OPENAI_DEPLOYMENT='gpt-35-turbo'
AZURE_OPENAI_EMBEDDINGS_DEPLOYMENT='text-embedding-ada-002'

Don't forget to save the .env file when done. You can now exit the file and return to the instructions for running the notebook.