Writing logs to /p/qdata/jm8wx/research/text_attacks/textattack/outputs/training/distilbert-base-uncased-glue:cola-2020-06-29-12:03/log.txt. Loading nlp dataset glue, subset cola, split train. Loading nlp dataset glue, subset cola, split validation. Loaded dataset. Found: 2 labels: ([0, 1]) Loading transformers AutoModelForSequenceClassification: distilbert-base-uncased Tokenizing training data. (len: 8551) Tokenizing eval data (len: 1043) Loaded data and tokenized in 16.85510802268982s Training model across 4 GPUs ***** Running training ***** Num examples = 8551 Batch size = 64 Max sequence length = 128 Num steps = 665 Num epochs = 5 Learning rate = 3e-05 Eval accuracy: 77.75647171620325% Best acc found. Saved model to /p/qdata/jm8wx/research/text_attacks/textattack/outputs/training/distilbert-base-uncased-glue:cola-2020-06-29-12:03/. Eval accuracy: 79.86577181208054% Best acc found. Saved model to /p/qdata/jm8wx/research/text_attacks/textattack/outputs/training/distilbert-base-uncased-glue:cola-2020-06-29-12:03/. Eval accuracy: 82.35858101629914% Best acc found. Saved model to /p/qdata/jm8wx/research/text_attacks/textattack/outputs/training/distilbert-base-uncased-glue:cola-2020-06-29-12:03/. Eval accuracy: 82.35858101629914% Eval accuracy: 82.16682646212847% Saved tokenizer to /p/qdata/jm8wx/research/text_attacks/textattack/outputs/training/distilbert-base-uncased-glue:cola-2020-06-29-12:03/. Wrote README to /p/qdata/jm8wx/research/text_attacks/textattack/outputs/training/distilbert-base-uncased-glue:cola-2020-06-29-12:03/README.md. Wrote training args to /p/qdata/jm8wx/research/text_attacks/textattack/outputs/training/distilbert-base-uncased-glue:cola-2020-06-29-12:03/train_args.json.