🤖 Fine-Tune Phi-3 on Custom Data

Complete guide to fine-tuning Microsoft's Phi-3 on your own data and deploying to production.

🚀 Quick Start

Google Colab (Recommended)

Click the badge above to open in Colab and run cell-by-cell.

🎥 Video Tutorial

Watch the complete walkthrough: YouTube Link

Timestamps:

• 00:00 - Intro
• 00:59 - Fine-Tuning Processes
• 03:20 - Colab Code Review
• 15:14 - Conclusion

📁 Repository Structure

• notebooks/: Interactive Jupyter/Colab notebooks
• data/: Sample dataset

📊 Data Format

Your training data should be in JSONL format:

{
  "messages": [
    {"role": "system", "content": "You are a helpful assistant."},
    {"role": "user", "content": "What is X?"},
    {"role": "assistant", "content": "X is..."}
  ]
}

See data/sample_training_data.jsonl for examples.

🤝 Contributing

Contributions, feedback, and suggestions are welcome!