Update log.txt
Browse files
log.txt
ADDED
|
@@ -0,0 +1,27 @@
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
| 1 |
+
Writing logs to /p/qdata/jm8wx/research/text_attacks/textattack/outputs/training/roberta-base-ag_news-2020-07-03-13:06/log.txt.
|
| 2 |
+
Loading [94mnlp[0m dataset [94mag_news[0m, split [94mtrain[0m.
|
| 3 |
+
Loading [94mnlp[0m dataset [94mag_news[0m, split [94mtest[0m.
|
| 4 |
+
Loaded dataset. Found: 4 labels: ([0, 1, 2, 3])
|
| 5 |
+
Loading transformers AutoModelForSequenceClassification: roberta-base
|
| 6 |
+
Tokenizing training data. (len: 120000)
|
| 7 |
+
Tokenizing eval data (len: 7600)
|
| 8 |
+
Loaded data and tokenized in 147.60910987854004s
|
| 9 |
+
Training model across 4 GPUs
|
| 10 |
+
***** Running training *****
|
| 11 |
+
Num examples = 120000
|
| 12 |
+
Batch size = 16
|
| 13 |
+
Max sequence length = 128
|
| 14 |
+
Num steps = 37500
|
| 15 |
+
Num epochs = 5
|
| 16 |
+
Learning rate = 5e-05
|
| 17 |
+
Eval accuracy: 93.71052631578948%
|
| 18 |
+
Best acc found. Saved model to /p/qdata/jm8wx/research/text_attacks/textattack/outputs/training/roberta-base-ag_news-2020-07-03-13:06/.
|
| 19 |
+
Eval accuracy: 93.6842105263158%
|
| 20 |
+
Eval accuracy: 94.61842105263158%
|
| 21 |
+
Best acc found. Saved model to /p/qdata/jm8wx/research/text_attacks/textattack/outputs/training/roberta-base-ag_news-2020-07-03-13:06/.
|
| 22 |
+
Eval accuracy: 94.5921052631579%
|
| 23 |
+
Eval accuracy: 94.69736842105263%
|
| 24 |
+
Best acc found. Saved model to /p/qdata/jm8wx/research/text_attacks/textattack/outputs/training/roberta-base-ag_news-2020-07-03-13:06/.
|
| 25 |
+
Saved tokenizer <textattack.models.tokenizers.auto_tokenizer.AutoTokenizer object at 0x7fcac6e3aee0> to /p/qdata/jm8wx/research/text_attacks/textattack/outputs/training/roberta-base-ag_news-2020-07-03-13:06/.
|
| 26 |
+
Wrote README to /p/qdata/jm8wx/research/text_attacks/textattack/outputs/training/roberta-base-ag_news-2020-07-03-13:06/README.md.
|
| 27 |
+
Wrote training args to /p/qdata/jm8wx/research/text_attacks/textattack/outputs/training/roberta-base-ag_news-2020-07-03-13:06/train_args.json.
|