-
Notifications
You must be signed in to change notification settings - Fork 154
/
Copy pathfinetuned_rte.log
296 lines (296 loc) · 14 KB
/
finetuned_rte.log
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
188
189
190
191
192
193
194
195
196
197
198
199
200
201
202
203
204
205
206
207
208
209
210
211
212
213
214
215
216
217
218
219
220
221
222
223
224
225
226
227
228
229
230
231
232
233
234
235
236
237
238
239
240
241
242
243
244
245
246
247
248
249
250
251
252
253
254
255
256
257
258
259
260
261
262
263
264
265
266
267
268
269
270
271
272
273
274
275
276
277
278
279
280
281
282
283
284
285
286
287
288
289
290
291
292
293
294
295
296
INFO:root:Namespace(accumulate=None, batch_size=32, dev_batch_size=8, epochs=3, gpu=True, log_interval=10, lr=2e-05, max_len=128, optimizer='bertadam', seed=2, task_name='RTE', test_batch_size=8, warmup_ratio=0.1)
[11:44:54] src/storage/storage.cc:108: Using GPUPooledRoundedStorageManager.
INFO:root:BERTClassifier(
(bert): BERTModel(
(encoder): BERTEncoder(
(dropout_layer): Dropout(p = 0.1, axes=())
(layer_norm): BERTLayerNorm(eps=1e-12, axis=-1, center=True, scale=True, in_channels=768)
(transformer_cells): HybridSequential(
(0): BERTEncoderCell(
(dropout_layer): Dropout(p = 0.1, axes=())
(attention_cell): MultiHeadAttentionCell(
(_base_cell): DotProductAttentionCell(
(_dropout_layer): Dropout(p = 0.1, axes=())
)
(proj_query): Dense(768 -> 768, linear)
(proj_key): Dense(768 -> 768, linear)
(proj_value): Dense(768 -> 768, linear)
)
(proj): Dense(768 -> 768, linear)
(ffn): BERTPositionwiseFFN(
(ffn_1): Dense(768 -> 3072, linear)
(activation): GELU()
(ffn_2): Dense(3072 -> 768, linear)
(dropout_layer): Dropout(p = 0.1, axes=())
(layer_norm): BERTLayerNorm(eps=1e-12, axis=-1, center=True, scale=True, in_channels=768)
)
(layer_norm): BERTLayerNorm(eps=1e-12, axis=-1, center=True, scale=True, in_channels=768)
)
(1): BERTEncoderCell(
(dropout_layer): Dropout(p = 0.1, axes=())
(attention_cell): MultiHeadAttentionCell(
(_base_cell): DotProductAttentionCell(
(_dropout_layer): Dropout(p = 0.1, axes=())
)
(proj_query): Dense(768 -> 768, linear)
(proj_key): Dense(768 -> 768, linear)
(proj_value): Dense(768 -> 768, linear)
)
(proj): Dense(768 -> 768, linear)
(ffn): BERTPositionwiseFFN(
(ffn_1): Dense(768 -> 3072, linear)
(activation): GELU()
(ffn_2): Dense(3072 -> 768, linear)
(dropout_layer): Dropout(p = 0.1, axes=())
(layer_norm): BERTLayerNorm(eps=1e-12, axis=-1, center=True, scale=True, in_channels=768)
)
(layer_norm): BERTLayerNorm(eps=1e-12, axis=-1, center=True, scale=True, in_channels=768)
)
(2): BERTEncoderCell(
(dropout_layer): Dropout(p = 0.1, axes=())
(attention_cell): MultiHeadAttentionCell(
(_base_cell): DotProductAttentionCell(
(_dropout_layer): Dropout(p = 0.1, axes=())
)
(proj_query): Dense(768 -> 768, linear)
(proj_key): Dense(768 -> 768, linear)
(proj_value): Dense(768 -> 768, linear)
)
(proj): Dense(768 -> 768, linear)
(ffn): BERTPositionwiseFFN(
(ffn_1): Dense(768 -> 3072, linear)
(activation): GELU()
(ffn_2): Dense(3072 -> 768, linear)
(dropout_layer): Dropout(p = 0.1, axes=())
(layer_norm): BERTLayerNorm(eps=1e-12, axis=-1, center=True, scale=True, in_channels=768)
)
(layer_norm): BERTLayerNorm(eps=1e-12, axis=-1, center=True, scale=True, in_channels=768)
)
(3): BERTEncoderCell(
(dropout_layer): Dropout(p = 0.1, axes=())
(attention_cell): MultiHeadAttentionCell(
(_base_cell): DotProductAttentionCell(
(_dropout_layer): Dropout(p = 0.1, axes=())
)
(proj_query): Dense(768 -> 768, linear)
(proj_key): Dense(768 -> 768, linear)
(proj_value): Dense(768 -> 768, linear)
)
(proj): Dense(768 -> 768, linear)
(ffn): BERTPositionwiseFFN(
(ffn_1): Dense(768 -> 3072, linear)
(activation): GELU()
(ffn_2): Dense(3072 -> 768, linear)
(dropout_layer): Dropout(p = 0.1, axes=())
(layer_norm): BERTLayerNorm(eps=1e-12, axis=-1, center=True, scale=True, in_channels=768)
)
(layer_norm): BERTLayerNorm(eps=1e-12, axis=-1, center=True, scale=True, in_channels=768)
)
(4): BERTEncoderCell(
(dropout_layer): Dropout(p = 0.1, axes=())
(attention_cell): MultiHeadAttentionCell(
(_base_cell): DotProductAttentionCell(
(_dropout_layer): Dropout(p = 0.1, axes=())
)
(proj_query): Dense(768 -> 768, linear)
(proj_key): Dense(768 -> 768, linear)
(proj_value): Dense(768 -> 768, linear)
)
(proj): Dense(768 -> 768, linear)
(ffn): BERTPositionwiseFFN(
(ffn_1): Dense(768 -> 3072, linear)
(activation): GELU()
(ffn_2): Dense(3072 -> 768, linear)
(dropout_layer): Dropout(p = 0.1, axes=())
(layer_norm): BERTLayerNorm(eps=1e-12, axis=-1, center=True, scale=True, in_channels=768)
)
(layer_norm): BERTLayerNorm(eps=1e-12, axis=-1, center=True, scale=True, in_channels=768)
)
(5): BERTEncoderCell(
(dropout_layer): Dropout(p = 0.1, axes=())
(attention_cell): MultiHeadAttentionCell(
(_base_cell): DotProductAttentionCell(
(_dropout_layer): Dropout(p = 0.1, axes=())
)
(proj_query): Dense(768 -> 768, linear)
(proj_key): Dense(768 -> 768, linear)
(proj_value): Dense(768 -> 768, linear)
)
(proj): Dense(768 -> 768, linear)
(ffn): BERTPositionwiseFFN(
(ffn_1): Dense(768 -> 3072, linear)
(activation): GELU()
(ffn_2): Dense(3072 -> 768, linear)
(dropout_layer): Dropout(p = 0.1, axes=())
(layer_norm): BERTLayerNorm(eps=1e-12, axis=-1, center=True, scale=True, in_channels=768)
)
(layer_norm): BERTLayerNorm(eps=1e-12, axis=-1, center=True, scale=True, in_channels=768)
)
(6): BERTEncoderCell(
(dropout_layer): Dropout(p = 0.1, axes=())
(attention_cell): MultiHeadAttentionCell(
(_base_cell): DotProductAttentionCell(
(_dropout_layer): Dropout(p = 0.1, axes=())
)
(proj_query): Dense(768 -> 768, linear)
(proj_key): Dense(768 -> 768, linear)
(proj_value): Dense(768 -> 768, linear)
)
(proj): Dense(768 -> 768, linear)
(ffn): BERTPositionwiseFFN(
(ffn_1): Dense(768 -> 3072, linear)
(activation): GELU()
(ffn_2): Dense(3072 -> 768, linear)
(dropout_layer): Dropout(p = 0.1, axes=())
(layer_norm): BERTLayerNorm(eps=1e-12, axis=-1, center=True, scale=True, in_channels=768)
)
(layer_norm): BERTLayerNorm(eps=1e-12, axis=-1, center=True, scale=True, in_channels=768)
)
(7): BERTEncoderCell(
(dropout_layer): Dropout(p = 0.1, axes=())
(attention_cell): MultiHeadAttentionCell(
(_base_cell): DotProductAttentionCell(
(_dropout_layer): Dropout(p = 0.1, axes=())
)
(proj_query): Dense(768 -> 768, linear)
(proj_key): Dense(768 -> 768, linear)
(proj_value): Dense(768 -> 768, linear)
)
(proj): Dense(768 -> 768, linear)
(ffn): BERTPositionwiseFFN(
(ffn_1): Dense(768 -> 3072, linear)
(activation): GELU()
(ffn_2): Dense(3072 -> 768, linear)
(dropout_layer): Dropout(p = 0.1, axes=())
(layer_norm): BERTLayerNorm(eps=1e-12, axis=-1, center=True, scale=True, in_channels=768)
)
(layer_norm): BERTLayerNorm(eps=1e-12, axis=-1, center=True, scale=True, in_channels=768)
)
(8): BERTEncoderCell(
(dropout_layer): Dropout(p = 0.1, axes=())
(attention_cell): MultiHeadAttentionCell(
(_base_cell): DotProductAttentionCell(
(_dropout_layer): Dropout(p = 0.1, axes=())
)
(proj_query): Dense(768 -> 768, linear)
(proj_key): Dense(768 -> 768, linear)
(proj_value): Dense(768 -> 768, linear)
)
(proj): Dense(768 -> 768, linear)
(ffn): BERTPositionwiseFFN(
(ffn_1): Dense(768 -> 3072, linear)
(activation): GELU()
(ffn_2): Dense(3072 -> 768, linear)
(dropout_layer): Dropout(p = 0.1, axes=())
(layer_norm): BERTLayerNorm(eps=1e-12, axis=-1, center=True, scale=True, in_channels=768)
)
(layer_norm): BERTLayerNorm(eps=1e-12, axis=-1, center=True, scale=True, in_channels=768)
)
(9): BERTEncoderCell(
(dropout_layer): Dropout(p = 0.1, axes=())
(attention_cell): MultiHeadAttentionCell(
(_base_cell): DotProductAttentionCell(
(_dropout_layer): Dropout(p = 0.1, axes=())
)
(proj_query): Dense(768 -> 768, linear)
(proj_key): Dense(768 -> 768, linear)
(proj_value): Dense(768 -> 768, linear)
)
(proj): Dense(768 -> 768, linear)
(ffn): BERTPositionwiseFFN(
(ffn_1): Dense(768 -> 3072, linear)
(activation): GELU()
(ffn_2): Dense(3072 -> 768, linear)
(dropout_layer): Dropout(p = 0.1, axes=())
(layer_norm): BERTLayerNorm(eps=1e-12, axis=-1, center=True, scale=True, in_channels=768)
)
(layer_norm): BERTLayerNorm(eps=1e-12, axis=-1, center=True, scale=True, in_channels=768)
)
(10): BERTEncoderCell(
(dropout_layer): Dropout(p = 0.1, axes=())
(attention_cell): MultiHeadAttentionCell(
(_base_cell): DotProductAttentionCell(
(_dropout_layer): Dropout(p = 0.1, axes=())
)
(proj_query): Dense(768 -> 768, linear)
(proj_key): Dense(768 -> 768, linear)
(proj_value): Dense(768 -> 768, linear)
)
(proj): Dense(768 -> 768, linear)
(ffn): BERTPositionwiseFFN(
(ffn_1): Dense(768 -> 3072, linear)
(activation): GELU()
(ffn_2): Dense(3072 -> 768, linear)
(dropout_layer): Dropout(p = 0.1, axes=())
(layer_norm): BERTLayerNorm(eps=1e-12, axis=-1, center=True, scale=True, in_channels=768)
)
(layer_norm): BERTLayerNorm(eps=1e-12, axis=-1, center=True, scale=True, in_channels=768)
)
(11): BERTEncoderCell(
(dropout_layer): Dropout(p = 0.1, axes=())
(attention_cell): MultiHeadAttentionCell(
(_base_cell): DotProductAttentionCell(
(_dropout_layer): Dropout(p = 0.1, axes=())
)
(proj_query): Dense(768 -> 768, linear)
(proj_key): Dense(768 -> 768, linear)
(proj_value): Dense(768 -> 768, linear)
)
(proj): Dense(768 -> 768, linear)
(ffn): BERTPositionwiseFFN(
(ffn_1): Dense(768 -> 3072, linear)
(activation): GELU()
(ffn_2): Dense(3072 -> 768, linear)
(dropout_layer): Dropout(p = 0.1, axes=())
(layer_norm): BERTLayerNorm(eps=1e-12, axis=-1, center=True, scale=True, in_channels=768)
)
(layer_norm): BERTLayerNorm(eps=1e-12, axis=-1, center=True, scale=True, in_channels=768)
)
)
)
(word_embed): HybridSequential(
(0): Embedding(30522 -> 768, float32)
(1): Dropout(p = 0.1, axes=())
)
(token_type_embed): HybridSequential(
(0): Embedding(2 -> 768, float32)
(1): Dropout(p = 0.1, axes=())
)
(pooler): Dense(768 -> 768, Activation(tanh))
)
(classifier): HybridSequential(
(0): Dropout(p = 0.1, axes=())
(1): Dense(None -> 2, linear)
)
)
INFO:root:Setting MXNET_GPU_MEM_POOL_TYPE="Round" may lead to higher memory usage and faster speed. If you encounter OOM errors, please unset this environment variable.
INFO:root:[Epoch 1 Batch 10/82] loss=0.7496, lr=0.0000087, metrics=accuracy:0.5217
INFO:root:[Epoch 1 Batch 20/82] loss=0.7023, lr=0.0000174, metrics=accuracy:0.5235
INFO:root:[Epoch 1 Batch 30/82] loss=0.6999, lr=0.0000193, metrics=accuracy:0.5207
INFO:root:[Epoch 1 Batch 40/82] loss=0.6685, lr=0.0000184, metrics=accuracy:0.5370
INFO:root:[Epoch 1 Batch 50/82] loss=0.6828, lr=0.0000174, metrics=accuracy:0.5459
INFO:root:[Epoch 1 Batch 60/82] loss=0.6819, lr=0.0000165, metrics=accuracy:0.5490
INFO:root:[Epoch 1 Batch 70/82] loss=0.6882, lr=0.0000155, metrics=accuracy:0.5545
INFO:root:[Epoch 1 Batch 80/82] loss=0.6419, lr=0.0000146, metrics=accuracy:0.5639
INFO:root:validation metrics:accuracy:0.6570
INFO:root:Time cost=47.0s
INFO:root:[Epoch 2 Batch 10/82] loss=0.5203, lr=0.0000134, metrics=accuracy:0.7625
INFO:root:[Epoch 2 Batch 20/82] loss=0.4785, lr=0.0000125, metrics=accuracy:0.7731
INFO:root:[Epoch 2 Batch 30/82] loss=0.4338, lr=0.0000115, metrics=accuracy:0.7928
INFO:root:[Epoch 2 Batch 40/82] loss=0.4582, lr=0.0000106, metrics=accuracy:0.7915
INFO:root:[Epoch 2 Batch 50/82] loss=0.4715, lr=0.0000096, metrics=accuracy:0.7950
INFO:root:[Epoch 2 Batch 60/82] loss=0.4851, lr=0.0000087, metrics=accuracy:0.7931
INFO:root:[Epoch 2 Batch 70/82] loss=0.4522, lr=0.0000077, metrics=accuracy:0.7903
INFO:root:[Epoch 2 Batch 80/82] loss=0.4203, lr=0.0000068, metrics=accuracy:0.7918
INFO:root:validation metrics:accuracy:0.6968
INFO:root:Time cost=45.1s
INFO:root:[Epoch 3 Batch 10/82] loss=0.2233, lr=0.0000056, metrics=accuracy:0.9529
INFO:root:[Epoch 3 Batch 20/82] loss=0.2608, lr=0.0000047, metrics=accuracy:0.9303
INFO:root:[Epoch 3 Batch 30/82] loss=0.2168, lr=0.0000037, metrics=accuracy:0.9306
INFO:root:[Epoch 3 Batch 40/82] loss=0.2667, lr=0.0000028, metrics=accuracy:0.9258
INFO:root:[Epoch 3 Batch 50/82] loss=0.1977, lr=0.0000018, metrics=accuracy:0.9282
INFO:root:[Epoch 3 Batch 60/82] loss=0.2957, lr=0.0000009, metrics=accuracy:0.9245
INFO:root:[Epoch 3 Batch 70/82] loss=0.2671, lr=-0.0000001, metrics=accuracy:0.9230
INFO:root:[Epoch 3 Batch 80/82] loss=0.3225, lr=-0.0000010, metrics=accuracy:0.9167
INFO:root:validation metrics:accuracy:0.7076
INFO:root:Time cost=45.4s