Introduction
Foundation models are pre-trained models that provide you with a great starting point. By using a foundation model, you can save time and effort as you need less data to train a model for your specific machine learning use case.
Imagine you're a data scientist working for a hotel booking agency. When customers browse through different hotels, one of the most important factors in deciding which hotel to book is reviews from other travelers.
As a data scientist, you may want to extract insights from the hotel reviews to find out why certain hotels are preferred over others. To extract information from hotel reviews, you can use Large Language Models (LLMs) that are designed for Natural Language Processing (NLP).
LLMs leverage deep learning techniques to understand and generate human language. Deep learning is a subfield of machine learning that involves training artificial neural networks with multiple layers to extract hierarchical patterns and representations from data. Training neural networks can be costly as it requires high volumes of data and powerful compute.
Instead of training your own LLM from scratch, you can use a pretrained model that you fine-tune using your own data. Imagine you want to detect sentiment in hotel reviews. You may want to categorize any newly posted reviews as describing the hotel as terrible, average, or excellent. You can use a small set of categorized hotel reviews to fine-tune a pretrained foundation model.
In this module, you learn how to fine-tune a foundation model from the model catalog in Azure Machine Learning.
Learning objectives
In this module, you learn how to:
- When to fine-tune a foundation model from the model catalog.
- Fine-tune a foundation model.
- Deploy and test a fine-tuned model.