Training vs Fine-tuning vs Prompting

Training vs Fine-tuning vs Prompting

Overview

• Training from scratch: requires large datasets and compute; rarely needed in capstone
• Fine-tuning: adapt a pre-trained model to your data
• Prompting: steer LLMs with instructions and examples; cheapest and fastest

When to use which

• Start with prompting or retrieval-augmented generation (RAG)
• Fine-tune if consistent task-specific behavior is needed

Checklist

• Measure baseline with prompting first
• Track datasets and hyperparameters for fine-tuning

Resources

• Hugging Face course
• OpenAI/Anthropic model guides