5fa1a76
1
2
We'll use the CoLA dataset from the GLUE benchmark, since it's a simple binary text classification task, and just take the training split for now.
