kaiser/distilling.out at master · machinereading/kaiser · GitHub

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
100
101
102
103
104
105
106
107
108
109
110
111
112
113
114
115
116
117
118
119
120
121
122
123
124
125
126
127
128
129
130
131
132
133
134
135
136
137
138
139
140
141
142
143
144
145
146
147
148
149
150
151
152
153
154
155
156
157
158
159
160
161
162
163
164
165
166
167
168
169
170
171
172
173
174
175
176
177
178
179
180
181
182
183
184
185
186
187
Using TensorFlow backend.
### Korean FrameNet ###
	# contact: hahmyg@kaist, hahmyg@gmail.com #


### loading Korean FrameNet 1.1 data...
	# of instances in training data: 17838
	# of instances in dev data: 2548
	# of instances in test data: 5097
# of instances in trn: 17838
# of instances in dev: 2548
# of instances in tst: 5097
data example: [['태풍', 'Hugo가', '남긴', '피해들과', '회사', '내', '몇몇', '주요', '부서들의', '저조한', '실적들을', '반영하여,', 'Aetna', 'Life', 'and', 'Casualty', 'Co.의', '3분기', '<tgt>', '순이익이', '</tgt>', '182.6', '백만', '달러', '또는', '주당', '1.63', '달러로', '22', '%', '하락하였다.'], ['_', '_', '_', '_', '_', '_', '_', '_', '_', '_', '_', '_', '_', '_', '_', '_', '_', '_', '_', '이익.n', '_', '_', '_', '_', '_', '_', '_', '_', '_', '_', '_'], ['_', '_', '_', '_', '_', '_', '_', '_', '_', '_', '_', '_', '_', '_', '_', '_', '_', '_', '_', 'Earnings_and_losses', '_', '_', '_', '_', '_', '_', '_', '_', '_', '_', '_'], ['O', 'O', 'O', 'O', 'O', 'O', 'O', 'O', 'O', 'O', 'O', 'O', 'B-Earner', 'I-Earner', 'I-Earner', 'I-Earner', 'I-Earner', 'B-Time', 'X', 'O', 'X', 'O', 'O', 'O', 'O', 'O', 'O', 'O', 'O', 'O', 'O']]

### TRAINING
MODEL: framenet
LANGUAGE: multi
PRETRAINED BERT: bert-base-multilingual-cased
training data:
	(ko): 17838
BATCH_SIZE: 3
MAX_LEN: 256

used dictionary:
	 /disk/kaiser/kaiser/src/../koreanframenet/resource/info/mul_lu2idx.json
	 /disk/kaiser/kaiser/src/../koreanframenet/resource/info/mul_lufrmap.json
	 /disk/kaiser/kaiser/src/../koreanframenet/resource/info/mul_bio_frargmap.json
teacher model: /disk/data/models/frameBERT/frameBERT_en/
original model: BERT-multilingual-base

	your model would be saved at /disk/data/models/framenet/distilling/

### converting data to BERT input...
	 ...is done: 0hour:0min:25sec
	#of instance: 17838 17838
Epoch:   0%|                                              | 0/50 [00:00<?, ?it/s]Train loss: 59.39154880920164

	your model is saved: /disk/data/models/framenet/distilling/0/
Epoch:   2%|▋                                | 1/50 [45:52<37:27:52, 2752.50s/it]Train loss: 37.024616148910496

	your model is saved: /disk/data/models/framenet/distilling/1/
Epoch:   4%|█▏                             | 2/50 [1:36:52<37:55:50, 2844.79s/it]Train loss: 31.80846098445298

	your model is saved: /disk/data/models/framenet/distilling/2/
Epoch:   6%|█▊                             | 3/50 [2:27:51<37:58:50, 2909.15s/it]Train loss: 27.795623297166873

	your model is saved: /disk/data/models/framenet/distilling/3/
Epoch:   8%|██▍                            | 4/50 [3:18:56<37:46:11, 2955.90s/it]Train loss: 24.731728677271676

	your model is saved: /disk/data/models/framenet/distilling/4/
Epoch:  10%|███                            | 5/50 [4:10:04<37:22:04, 2989.43s/it]Train loss: 22.60041129793577

	your model is saved: /disk/data/models/framenet/distilling/5/
Epoch:  12%|███▋                           | 6/50 [5:01:05<36:48:02, 3010.97s/it]Train loss: 20.778911683601923

	your model is saved: /disk/data/models/framenet/distilling/6/
Epoch:  14%|████▎                          | 7/50 [5:52:15<36:10:28, 3028.56s/it]Train loss: 19.392256444446012

	your model is saved: /disk/data/models/framenet/distilling/7/
Epoch:  16%|████▉                          | 8/50 [6:43:25<35:28:41, 3041.00s/it]Train loss: 18.167695532172285

	your model is saved: /disk/data/models/framenet/distilling/8/
Epoch:  18%|█████▌                         | 9/50 [7:34:28<34:42:32, 3047.62s/it]Train loss: 17.159835035537093

	your model is saved: /disk/data/models/framenet/distilling/9/
Epoch:  20%|██████                        | 10/50 [8:25:38<33:56:13, 3054.34s/it]Train loss: 16.31597974546582

	your model is saved: /disk/data/models/framenet/distilling/10/
Epoch:  22%|██████▌                       | 11/50 [9:16:53<33:09:21, 3060.54s/it]Train loss: 15.468475778456533

	your model is saved: /disk/data/models/framenet/distilling/11/
Epoch:  24%|██████▉                      | 12/50 [10:07:59<32:19:20, 3062.12s/it]Train loss: 14.752258084376091

	your model is saved: /disk/data/models/framenet/distilling/12/
Epoch:  26%|███████▌                     | 13/50 [10:59:12<31:30:20, 3065.41s/it]Train loss: 14.113097571500656

	your model is saved: /disk/data/models/framenet/distilling/13/
Epoch:  28%|████████                     | 14/50 [11:50:40<30:43:24, 3072.34s/it]Train loss: 13.589403949720548

	your model is saved: /disk/data/models/framenet/distilling/14/
Epoch:  30%|████████▋                    | 15/50 [12:41:57<29:52:53, 3073.54s/it]Train loss: 13.000793678408474

	your model is saved: /disk/data/models/framenet/distilling/15/
Epoch:  32%|█████████▎                   | 16/50 [13:33:18<29:02:55, 3075.74s/it]Train loss: 12.557038174347568

	your model is saved: /disk/data/models/framenet/distilling/16/
Epoch:  34%|█████████▊                   | 17/50 [14:24:53<28:14:56, 3081.70s/it]Train loss: 12.063018780503656

	your model is saved: /disk/data/models/framenet/distilling/17/
Epoch:  36%|██████████▍                  | 18/50 [15:16:25<27:25:10, 3084.70s/it]Train loss: 11.721391938087958

	your model is saved: /disk/data/models/framenet/distilling/18/
Epoch:  38%|███████████                  | 19/50 [16:07:43<26:32:39, 3082.56s/it]Train loss: 11.309914528220903

	your model is saved: /disk/data/models/framenet/distilling/19/
Epoch:  40%|███████████▌                 | 20/50 [16:59:00<25:40:35, 3081.17s/it]Train loss: 10.97218985470144

	your model is saved: /disk/data/models/framenet/distilling/20/
Epoch:  42%|████████████▏                | 21/50 [17:50:18<24:48:39, 3079.98s/it]Train loss: 10.626068878526684

	your model is saved: /disk/data/models/framenet/distilling/21/
Epoch:  44%|████████████▊                | 22/50 [18:41:32<23:56:34, 3078.36s/it]Train loss: 10.335406667769977

	your model is saved: /disk/data/models/framenet/distilling/22/
Epoch:  46%|█████████████▎               | 23/50 [19:32:51<23:05:16, 3078.39s/it]Train loss: 10.050212201428982

	your model is saved: /disk/data/models/framenet/distilling/23/
Epoch:  48%|█████████████▉               | 24/50 [20:24:13<22:14:31, 3079.65s/it]Train loss: 9.803551743016914

	your model is saved: /disk/data/models/framenet/distilling/24/
Epoch:  50%|██████████████▌              | 25/50 [21:15:27<21:22:23, 3077.74s/it]Train loss: 9.534623263341748

	your model is saved: /disk/data/models/framenet/distilling/25/
Epoch:  52%|███████████████              | 26/50 [22:06:52<20:32:02, 3080.09s/it]Train loss: 9.30669885895848

	your model is saved: /disk/data/models/framenet/distilling/26/
Epoch:  54%|███████████████▋             | 27/50 [22:58:24<19:42:03, 3083.62s/it]Train loss: 9.06872954847353

	your model is saved: /disk/data/models/framenet/distilling/27/
Epoch:  56%|████████████████▏            | 28/50 [23:49:48<18:50:39, 3083.63s/it]Train loss: 8.855293762190671

	your model is saved: /disk/data/models/framenet/distilling/28/
Epoch:  58%|████████████████▊            | 29/50 [24:41:17<17:59:50, 3085.26s/it]Train loss: 8.648713634064652

	your model is saved: /disk/data/models/framenet/distilling/29/
Epoch:  60%|█████████████████▍           | 30/50 [25:32:55<17:09:40, 3089.03s/it]Train loss: 8.464723509407749

	your model is saved: /disk/data/models/framenet/distilling/30/
Epoch:  62%|█████████████████▉           | 31/50 [26:24:12<16:17:04, 3085.51s/it]Train loss: 8.28572597862132

	your model is saved: /disk/data/models/framenet/distilling/31/
Epoch:  64%|██████████████████▌          | 32/50 [27:15:30<15:24:59, 3083.33s/it]Train loss: 8.12280644766618

	your model is saved: /disk/data/models/framenet/distilling/32/
Epoch:  66%|███████████████████▏         | 33/50 [28:06:54<14:33:38, 3083.47s/it]Train loss: 7.93051823172712

	your model is saved: /disk/data/models/framenet/distilling/33/
Epoch:  68%|███████████████████▋         | 34/50 [28:58:08<13:41:30, 3080.66s/it]Train loss: 7.785495751941801

	your model is saved: /disk/data/models/framenet/distilling/34/
Epoch:  70%|████████████████████▎        | 35/50 [29:49:20<12:49:32, 3078.16s/it]Train loss: 7.650097507100052

	your model is saved: /disk/data/models/framenet/distilling/35/
Epoch:  72%|████████████████████▉        | 36/50 [30:40:32<11:57:45, 3076.14s/it]Train loss: 7.51084882740697

	your model is saved: /disk/data/models/framenet/distilling/36/
Epoch:  74%|█████████████████████▍       | 37/50 [31:31:43<11:06:09, 3074.61s/it]Train loss: 7.351098593180323

	your model is saved: /disk/data/models/framenet/distilling/37/
Epoch:  76%|██████████████████████       | 38/50 [32:22:46<10:14:15, 3071.31s/it]Train loss: 7.244942439773289

	your model is saved: /disk/data/models/framenet/distilling/38/
Epoch:  78%|███████████████████████▍      | 39/50 [33:13:38<9:21:59, 3065.40s/it]Train loss: 7.11489543869084

	your model is saved: /disk/data/models/framenet/distilling/39/
Epoch:  80%|████████████████████████      | 40/50 [34:04:14<8:29:26, 3056.61s/it]Train loss: 6.9898716030279635

	your model is saved: /disk/data/models/framenet/distilling/40/
Epoch:  82%|████████████████████████▌     | 41/50 [34:54:59<7:37:57, 3053.07s/it]Train loss: 6.890018276882781

	your model is saved: /disk/data/models/framenet/distilling/41/
Epoch:  84%|█████████████████████████▏    | 42/50 [35:45:44<6:46:44, 3050.61s/it]Train loss: 6.788874380804255

	your model is saved: /disk/data/models/framenet/distilling/42/
Epoch:  86%|█████████████████████████▊    | 43/50 [36:36:28<5:55:40, 3048.67s/it]Train loss: 6.683639158093484

	your model is saved: /disk/data/models/framenet/distilling/43/
Epoch:  88%|██████████████████████████▍   | 44/50 [37:27:16<5:04:50, 3048.36s/it]Train loss: 6.5730640079451454

	your model is saved: /disk/data/models/framenet/distilling/44/
Epoch:  90%|███████████████████████████   | 45/50 [38:18:04<4:14:01, 3048.38s/it]Train loss: 6.467643125086091

	your model is saved: /disk/data/models/framenet/distilling/45/
Epoch:  92%|███████████████████████████▌  | 46/50 [39:08:18<3:22:32, 3038.09s/it]Train loss: 6.358480372975821

	your model is saved: /disk/data/models/framenet/distilling/46/
Epoch:  94%|████████████████████████████▏ | 47/50 [39:59:04<2:32:01, 3040.36s/it]Train loss: 6.30173086213857

	your model is saved: /disk/data/models/framenet/distilling/47/
Epoch:  96%|████████████████████████████▊ | 48/50 [40:49:49<1:41:23, 3041.69s/it]Train loss: 6.202840307947365

	your model is saved: /disk/data/models/framenet/distilling/48/
Epoch:  98%|███████████████████████████████▎| 49/50 [41:40:36<50:43, 3043.45s/it]Train loss: 6.122672133641415

	your model is saved: /disk/data/models/framenet/distilling/49/
Epoch: 100%|████████████████████████████████| 50/50 [42:31:20<00:00, 3043.47s/it]Epoch: 100%|████████████████████████████████| 50/50 [42:31:20<00:00, 3061.60s/it]
...training is done