Writing logs to /p/qdata/jm8wx/research/text_attacks/textattack/outputs/training/albert-base-v2-ag_news-2020-07-01-02:15/log.txt.
Loading nlp dataset ag_news, split train.
Loading nlp dataset ag_news, split test.
Loaded dataset. Found: 4 labels: ([0, 1, 2, 3])
Loading transformers AutoModelForSequenceClassification: albert-base-v2
Tokenizing training data. (len: 120000)
Tokenizing eval data (len: 7600)
Loaded data and tokenized in 165.6598937511444s
Training model across 4 GPUs
***** Running training *****
	Num examples = 120000
	Batch size = 16
	Max sequence length = 128
	Num steps = 37500
	Num epochs = 5
	Learning rate = 2e-05
Eval accuracy: 92.97368421052632%
Best acc found. Saved model to /p/qdata/jm8wx/research/text_attacks/textattack/outputs/training/albert-base-v2-ag_news-2020-07-01-02:15/.
Eval accuracy: 94.03947368421053%
Best acc found. Saved model to /p/qdata/jm8wx/research/text_attacks/textattack/outputs/training/albert-base-v2-ag_news-2020-07-01-02:15/.
Eval accuracy: 94.35526315789474%
Best acc found. Saved model to /p/qdata/jm8wx/research/text_attacks/textattack/outputs/training/albert-base-v2-ag_news-2020-07-01-02:15/.
Eval accuracy: 94.71052631578948%
Best acc found. Saved model to /p/qdata/jm8wx/research/text_attacks/textattack/outputs/training/albert-base-v2-ag_news-2020-07-01-02:15/.
Eval accuracy: 94.47368421052632%
Saved tokenizer <textattack.models.tokenizers.auto_tokenizer.AutoTokenizer object at 0x7fe892860880> to /p/qdata/jm8wx/research/text_attacks/textattack/outputs/training/albert-base-v2-ag_news-2020-07-01-02:15/.
Wrote README to /p/qdata/jm8wx/research/text_attacks/textattack/outputs/training/albert-base-v2-ag_news-2020-07-01-02:15/README.md.
Wrote training args to /p/qdata/jm8wx/research/text_attacks/textattack/outputs/training/albert-base-v2-ag_news-2020-07-01-02:15/train_args.json.