File size: 4,856 Bytes
082c3ef
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
1
2
3
4
5
6
7
8
9
10
11
12
13
14
15
16
17
18
19
20
21
22
23
24
25
26
27
28
29
30
31
32
33
34
35
36
37
38
39
40
41
42
43
44
45
46
47
48
49
50
51
52
53
54
55
56
57
58
59
60
61
62
63
64
65
66
67
68
69
70
71
72
73
74
75
76
77
78
79
80
81
82
83
84
85
86
87
88
89
90
91
92
93
94
95
96
97
98
99
Writing logs to /home/ubuntu/buildsCodes/Adversarial_training/trained_models/Multi-delete-our_bert-base-uncased-IMDB/train_log.txt.
Wrote original training args to /home/ubuntu/buildsCodes/Adversarial_training/trained_models/Multi-delete-our_bert-base-uncased-IMDB/training_args.json.
***** Running training *****
  Num examples = 25000
  Num epochs = 5
  Num clean epochs = 1
  Instantaneous batch size per device = 8
  Total train batch size (w. parallel, distributed & accumulation) = 32
  Gradient accumulation steps = 4
  Total optimization steps = 4410
==========================================================
Epoch 1
Running clean epoch 1/1
Writing logs to /home/ubuntu/buildsCodes/Adversarial_training/trained_models/Multi-delete-our_bert-base-uncased-IMDB/train_log.txt.
Wrote original training args to /home/ubuntu/buildsCodes/Adversarial_training/trained_models/Multi-delete-our_bert-base-uncased-IMDB/training_args.json.
***** Running training *****
  Num examples = 25000
  Num epochs = 5
  Num clean epochs = 1
  Instantaneous batch size per device = 8
  Total train batch size (w. parallel, distributed & accumulation) = 32
  Gradient accumulation steps = 4
  Total optimization steps = 4410
==========================================================
Epoch 1
Running clean epoch 1/1
Writing logs to /home/ubuntu/buildsCodes/Adversarial_training/trained_models/Multi-delete-our_bert-base-uncased-IMDB/train_log.txt.
Wrote original training args to /home/ubuntu/buildsCodes/Adversarial_training/trained_models/Multi-delete-our_bert-base-uncased-IMDB/training_args.json.
***** Running training *****
  Num examples = 25000
  Num epochs = 5
  Num clean epochs = 1
  Instantaneous batch size per device = 8
  Total train batch size (w. parallel, distributed & accumulation) = 32
  Gradient accumulation steps = 4
  Total optimization steps = 4410
==========================================================
Epoch 1
Running clean epoch 1/1
Writing logs to /home/ubuntu/buildsCodes/Adversarial_training/trained_models/Multi-delete-our_bert-base-uncased-IMDB/train_log.txt.
Wrote original training args to /home/ubuntu/buildsCodes/Adversarial_training/trained_models/Multi-delete-our_bert-base-uncased-IMDB/training_args.json.
***** Running training *****
  Num examples = 25000
  Num epochs = 5
  Num clean epochs = 1
  Instantaneous batch size per device = 8
  Total train batch size (w. parallel, distributed & accumulation) = 32
  Gradient accumulation steps = 4
  Total optimization steps = 4410
==========================================================
Epoch 1
Running clean epoch 1/1
Train accuracy: 97.48%
Eval accuracy: 90.31%
Best score found. Saved model to /home/ubuntu/buildsCodes/Adversarial_training/trained_models/Multi-delete-our_bert-base-uncased-IMDB//best_model/
==========================================================
Epoch 2
Attacking model to generate new adversarial training set...
Total number of attack results: 4403
Attack success rate: 91.43% [4000 / 4375]
Train accuracy: 98.84%
Eval accuracy: 93.46%
Best score found. Saved model to /home/ubuntu/buildsCodes/Adversarial_training/trained_models/Multi-delete-our_bert-base-uncased-IMDB//best_model/
==========================================================
Epoch 3
Attacking model to generate new adversarial training set...
Writing logs to /home/ubuntu/buildsCodes/Adversarial_training/trained_models/Multi-delete-our_bert-base-uncased-IMDB/train_log.txt.
Wrote original training args to /home/ubuntu/buildsCodes/Adversarial_training/trained_models/Multi-delete-our_bert-base-uncased-IMDB/training_args.json.
***** Running training *****
  Num examples = 25000
  Num epochs = 5
  Num clean epochs = 1
  Instantaneous batch size per device = 8
  Total train batch size (w. parallel, distributed & accumulation) = 32
  Gradient accumulation steps = 4
  Total optimization steps = 4410
==========================================================
Epoch 1
Running clean epoch 1/1
Train accuracy: 97.48%
Eval accuracy: 90.31%
Best score found. Saved model to /home/ubuntu/buildsCodes/Adversarial_training/trained_models/Multi-delete-our_bert-base-uncased-IMDB//best_model/
==========================================================
Epoch 2
Attacking model to generate new adversarial training set...
Train accuracy: 98.89%
Eval accuracy: 93.25%
Best score found. Saved model to /home/ubuntu/buildsCodes/Adversarial_training/trained_models/Multi-delete-our_bert-base-uncased-IMDB//best_model/
==========================================================
Epoch 3
Attacking model to generate new adversarial training set...
Total number of attack results: 6088
Attack success rate: 65.77% [4000 / 6082]
Train accuracy: 70.22%
Eval accuracy: 93.25%
==========================================================
Epoch 4
Attacking model to generate new adversarial training set...