Tutorial on using T5 model for text summarization

[pmabbo13]

Description

Add a tutorial which demonstrates how to use the T5 model for text summarization.

Process

Demonstrate the end-to-end process for generating text summaries of articles from the CNNDM dataset using TorchText's pre-trained T5 model with base configuration. This involved showcasing how to:

1. Initialize the text pre-processing pipeline
2. Load the CNNDM dataset and process its text to include the task prefix needed for the model
3. Create a generator that iteratively calls on the decoder in order to continuously expand the output sequences until all sequences in the batch generate the end-of-sequence token. A greedy search is used to predict the next word in each sequence
4. Load the pre-trained T5 model and call on the generator to produce summaries for a small batch of articles taken from the CNNDM test set

Test

Run BUILD_GALLERY=1 make 'SPHINXOPTS=-W' html in docs and review rendered document in docs/build/html/tutorials/cnndm_summarization.html

Follow-Up

• Implement a beam search generator Updating T5 demo to use beam search for generator #1869

[parmeet]

Overall, LGTM! Thank you @pmabbo13 for adding this end-2-end tutorial, this would be quite helpful for users to on-board T5 model. Let's make sure the tutorial can be build locally before merge :).

[Nayef211]

LGTM! Great work on adding a descriptive tutorial that is very easy to follow!