textattack/bert-base-uncased-RTE
34.1k
1Writing logs to /p/qdata/jm8wx/research/text_attacks/textattack/outputs/training/bert-base-uncased-glue:rte-2020-06-29-14:24/log.txt.2Loading [94mnlp[0m dataset [94mglue[0m, subset [94mrte[0m, split [94mtrain[0m.3Loading [94mnlp[0m dataset [94mglue[0m, subset [94mrte[0m, split [94mvalidation[0m.4Loaded dataset. Found: 2 labels: ([0, 1])5Loading transformers AutoModelForSequenceClassification: bert-base-uncased6Tokenizing training data. (len: 2490)7Tokenizing eval data (len: 277)8Loaded data and tokenized in 14.295648097991943s9Training model across 1 GPUs10***** Running training *****11 Num examples = 249012 Batch size = 12813 Max sequence length = 12814 Num steps = 9515 Num epochs = 516 Learning rate = 3e-0517Failed to predict with model <class 'transformers.modeling_bert.BertForSequenceClassification'>. Check tokenizer configuration.18Writing logs to /p/qdata/jm8wx/research/text_attacks/textattack/outputs/training/bert-base-uncased-glue:rte-2020-06-29-14:24/log.txt.19Loading [94mnlp[0m dataset [94mglue[0m, subset [94mrte[0m, split [94mtrain[0m.20Loading [94mnlp[0m dataset [94mglue[0m, subset [94mrte[0m, split [94mvalidation[0m.21Loaded dataset. Found: 2 labels: ([0, 1])22Loading transformers AutoModelForSequenceClassification: bert-base-uncased23Tokenizing training data. (len: 2490)24Tokenizing eval data (len: 277)25Loaded data and tokenized in 13.395596742630005s26Training model across 1 GPUs27***** Running training *****28 Num examples = 249029 Batch size = 830 Max sequence length = 12831 Num steps = 155532 Num epochs = 533 Learning rate = 2e-0534Eval accuracy: 68.23104693140795%35Best acc found. Saved model to /p/qdata/jm8wx/research/text_attacks/textattack/outputs/training/bert-base-uncased-glue:rte-2020-06-29-14:24/.36Eval accuracy: 70.03610108303249%37Best acc found. Saved model to /p/qdata/jm8wx/research/text_attacks/textattack/outputs/training/bert-base-uncased-glue:rte-2020-06-29-14:24/.38Eval accuracy: 72.56317689530685%39Best acc found. Saved model to /p/qdata/jm8wx/research/text_attacks/textattack/outputs/training/bert-base-uncased-glue:rte-2020-06-29-14:24/.40Eval accuracy: 69.67509025270758%41Eval accuracy: 69.31407942238266%42Saved tokenizer <textattack.models.tokenizers.auto_tokenizer.AutoTokenizer object at 0x7fc688911c10> to /p/qdata/jm8wx/research/text_attacks/textattack/outputs/training/bert-base-uncased-glue:rte-2020-06-29-14:24/.43Wrote README to /p/qdata/jm8wx/research/text_attacks/textattack/outputs/training/bert-base-uncased-glue:rte-2020-06-29-14:24/README.md.44Wrote training args to /p/qdata/jm8wx/research/text_attacks/textattack/outputs/training/bert-base-uncased-glue:rte-2020-06-29-14:24/train_args.json.45