Writing logs to /p/qdata/jm8wx/research/text_attacks/textattack/outputs/training/albert-base-v2-snli-2020-07-02-13:00/log.txt.
Loading nlp dataset snli, split train.
Loading nlp dataset snli, split validation.
Filtering samples with labels outside of [0, 1, 2].
Filtered 550152 train samples to 549367 points.
Filtered 10000 dev samples to 9842 points.
Loaded dataset. Found: 3 labels: ([0, 1, 2])
Loading transformers AutoModelForSequenceClassification: albert-base-v2
Tokenizing training data. (len: 549367)
Tokenizing eval data (len: 9842)
Loaded data and tokenized in 770.6083040237427s
Training model across 1 GPUs
***** Running training *****
	Num examples = 549367
	Batch size = 64
	Max sequence length = 64
	Num steps = 42915
	Num epochs = 5
	Learning rate = 2e-05
Eval accuracy: 89.86994513310303%
Best acc found. Saved model to /p/qdata/jm8wx/research/text_attacks/textattack/outputs/training/albert-base-v2-snli-2020-07-02-13:00/.
Eval accuracy: 89.8191424507214%
Eval accuracy: 90.6015037593985%
Best acc found. Saved model to /p/qdata/jm8wx/research/text_attacks/textattack/outputs/training/albert-base-v2-snli-2020-07-02-13:00/.
Eval accuracy: 90.27636659215607%
Eval accuracy: 90.25604551920341%
Saved tokenizer <textattack.models.tokenizers.auto_tokenizer.AutoTokenizer object at 0x7fd0840a5190> to /p/qdata/jm8wx/research/text_attacks/textattack/outputs/training/albert-base-v2-snli-2020-07-02-13:00/.
Wrote README to /p/qdata/jm8wx/research/text_attacks/textattack/outputs/training/albert-base-v2-snli-2020-07-02-13:00/README.md.
Wrote training args to /p/qdata/jm8wx/research/text_attacks/textattack/outputs/training/albert-base-v2-snli-2020-07-02-13:00/train_args.json.