Writing logs to /p/qdata/jm8wx/research/text_attacks/textattack/outputs/training/albert-base-v2-glue:sst2-2020-06-30-00:28/log.txt. Loading nlp dataset glue, subset sst2, split train. Loading nlp dataset glue, subset sst2, split validation. Loaded dataset. Found: 2 labels: ([0, 1]) Loading transformers AutoModelForSequenceClassification: albert-base-v2 Tokenizing training data. (len: 67349) Tokenizing eval data (len: 872) Loaded data and tokenized in 100.11767911911011s Training model across 1 GPUs ***** Running training ***** Num examples = 67349 Batch size = 32 Max sequence length = 64 Num steps = 10520 Num epochs = 5 Learning rate = 3e-05 Eval accuracy: 91.74311926605505% Best acc found. Saved model to /p/qdata/jm8wx/research/text_attacks/textattack/outputs/training/albert-base-v2-glue:sst2-2020-06-30-00:28/. Eval accuracy: 91.74311926605505% Eval accuracy: 92.54587155963303% Best acc found. Saved model to /p/qdata/jm8wx/research/text_attacks/textattack/outputs/training/albert-base-v2-glue:sst2-2020-06-30-00:28/. Eval accuracy: 91.62844036697247% Eval accuracy: 91.97247706422019% Saved tokenizer to /p/qdata/jm8wx/research/text_attacks/textattack/outputs/training/albert-base-v2-glue:sst2-2020-06-30-00:28/. Wrote README to /p/qdata/jm8wx/research/text_attacks/textattack/outputs/training/albert-base-v2-glue:sst2-2020-06-30-00:28/README.md. Wrote training args to /p/qdata/jm8wx/research/text_attacks/textattack/outputs/training/albert-base-v2-glue:sst2-2020-06-30-00:28/train_args.json.