202405032345
Status: #idea
Tags: Natural Language Processing (NLP)

Universal Language Model Fine-tuning (ULMFit

This is brilliant technique that consists in fine-tuning a pretrained language model say BERT on a dataset related to the task that you want to make it do, before actually training it for the task.

As a result the model is able to see more data, and the results are significantly better.

References

Practical Deep Learning for Coders#Chapter 10