Writing logs to /p/qdata/jm8wx/research/text_attacks/textattack/outputs/training/roberta-base-imdb-2020-06-30-17:56/log.txt. Loading nlp dataset imdb, split train. Loading nlp dataset imdb, split test. Loaded dataset. Found: 2 labels: ([0, 1]) Loading transformers AutoModelForSequenceClassification: roberta-base Tokenizing training data. (len: 25000) Tokenizing eval data (len: 25000) Loaded data and tokenized in 79.24297642707825s Training model across 4 GPUs ***** Running training ***** Num examples = 25000 Batch size = 64 Max sequence length = 128 Num steps = 1950 Num epochs = 5 Learning rate = 3e-05 Eval accuracy: 90.776% Best acc found. Saved model to /p/qdata/jm8wx/research/text_attacks/textattack/outputs/training/roberta-base-imdb-2020-06-30-17:56/. Eval accuracy: 91.35600000000001% Best acc found. Saved model to /p/qdata/jm8wx/research/text_attacks/textattack/outputs/training/roberta-base-imdb-2020-06-30-17:56/. Eval accuracy: 91.43599999999999% Best acc found. Saved model to /p/qdata/jm8wx/research/text_attacks/textattack/outputs/training/roberta-base-imdb-2020-06-30-17:56/. Eval accuracy: 91.172% Eval accuracy: 91.408% Saved tokenizer to /p/qdata/jm8wx/research/text_attacks/textattack/outputs/training/roberta-base-imdb-2020-06-30-17:56/. Wrote README to /p/qdata/jm8wx/research/text_attacks/textattack/outputs/training/roberta-base-imdb-2020-06-30-17:56/README.md. Wrote training args to /p/qdata/jm8wx/research/text_attacks/textattack/outputs/training/roberta-base-imdb-2020-06-30-17:56/train_args.json.