# Transformers for SAG

Train and evaluate transformer-based models for Short-Answer Grading

Use the script as 
`python run_SAG.py --do_eval --data_dir data/glue_data/SST-2 --model_type bert --model_name_or_path 'bert-base-uncased' --task_name sst-2 --output_dir output`

(This will take around four minutes to run on a CPU, but will be a lot quicker on a GPU...)