Uploading Custom Data

Let's do it.

The Tuning Workflow

1. Navigate: In AI Studio, click "New Tuned Model" (left sidebar).
2. Dataset: Click "Import" -> "Google Sheets" or "Upload CSV".
   • Preview the data to ensure columns mapped correctly (Input vs Output).
3. Configuration:
   • Base Model: Choose gemini-1.5-flash-001-tuning. (Not all versions are tunable).
   • Name: my-company-tone-v1.
4. Hyperparameters: Leave defaults or tweak Epochs.
5. Train: Click the blue button.

Monitoring

Tuning is not instant. It can take 10 minutes to 2 hours depending on data size.

• Status: You will see a progress bar.
• Can I close the tab?: Yes. It runs on Google Cloud. You will get an email (or see the status change) when done.

Accessing the Model

Once done, the model appears in your My Library. To use it in code:

model = genai.GenerativeModel('tunedModels/my-company-tone-v1')
response = model.generate_content("Hi")

Summary

The UI is simple. The hard part was preparing the CSV (Lesson 2). The easy part is clicking "Train."

In the final lesson of this module, we discuss Evaluating Fine-Tuned Models.