This is the code base for ANLP HUJI course exercise 1, fine tuning pretrained models to perform sentiment analysis on the SST2 dataset.
pip install -r requirements.txt
Run:
python ex1.py --max_train_samples <number of train samples> --max_eval_samples <number of validation samples> --max_predict_samples <number of prediction samples> --lr <learning rate> --num_train_epochs <number of training epochs> --batch_size <batch size> --do_train/--do_predict --model_path <path to prediction model>
If you use --do_predict, a prediction.txt file will be generated, containing prediction results for all test samples.