README.md

QuickFormer - SimpleTransformers, but even simpler

Quickly train and evaluate a Transformer classification model on a given dataset.

QuickFormer allows you to:

• Produce and evaluate a Transformer classification model by only one line of code
• Evaluate the model and find precision, recall, and F1 score for all classes
• Generate confusion matrices for the model

How to run

1. Prepare your initial data in a CSV file with the columns text and cat_label. Name the file model_name_input.csv where model_name is an arbitrary name.

2. Call the quickform() function. The only mandatory argument is the model name (model_name). The function will automatically load the data from the CSV file, train the model, evaluate it and saves all the results in files starting with model_name.

That's it!

Automatically generated files after each run

1. model_name_train_data.csv - the data used for training
2. model_name_test_data.csv - the data used for testing
3. model_name_confusion_matrix.png - a visualization of the confusion matrix
4. model_name_confusion_matrix_normalized.png - a visualization of the confusion matrix with normalized values
5. model_name_confusion_matrix.txt - the raw data of the confusion matrix
6. model_name_precision_recall_f1.txt - the data for precision, recall, and F1 score for each of the classes
7. model_name_codes.txt - relation between the codes and the different classes in the dataset
8. outputs_model_name - saved model, which can be later reused for inference

Enjoy using QuickFormer!