Tutorial shows how to train Text Summarization Models using Lightning Flash on the XSum Dataset
This example covers an abstractive Text Summarization deep learning task
What is Text Summarization
Training the model using train.py script
Loading Grid Weights in Flash and Running Summarization Model
This tutorial uses PyTorch Lightning
Task: Text Summarization
Text Summarization is the task of generating a consise overview of a text's main point into a short sentence/description. For example, taking a web article and describing the topic in a short sentence.
The Extreme Summarization (XSum) dataset is a dataset for evaluation of abstractive single-document summarization systems consisting of 226,711 news articles collected from BBC (2010 to 2017) accompanied with a one-sentence summary. The articles cover a wide variety of subjects (e.g., News, Politics, Sports, Weather, Business, Technology, Science, Health, Family, Education, Entertainment and Arts)
Step 1: Model
PyTorch Lightning Flash enables the quick training, fine tuning, and inferencing of SOTA object detection algorithms such as RetinaNet.
For this demo, we're going to be using the code here