-
Notifications
You must be signed in to change notification settings - Fork 1
/
Copy pathstartup_4
258 lines (257 loc) · 39.6 KB
/
startup_4
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
{ // block 0
persist trainable param embedding_0.w_0 : LOD_TENSOR.shape(512, 768).dtype(FP32).stop_gradient(False)
persist trainable param embedding_1.w_0 : LOD_TENSOR.shape(2, 768).dtype(FP32).stop_gradient(False)
persist trainable param layer_norm_0.w_0 : LOD_TENSOR.shape(768,).dtype(FP32).stop_gradient(False)
persist trainable param layer_norm_0.b_0 : LOD_TENSOR.shape(768,).dtype(FP32).stop_gradient(False)
persist trainable param linear_0.w_0 : LOD_TENSOR.shape(768, 768).dtype(FP32).stop_gradient(False)
persist trainable param linear_0.b_0 : LOD_TENSOR.shape(768,).dtype(FP32).stop_gradient(False)
persist trainable param linear_1.w_0 : LOD_TENSOR.shape(768, 768).dtype(FP32).stop_gradient(False)
persist trainable param layer_norm_25.w_0 : LOD_TENSOR.shape(768,).dtype(FP32).stop_gradient(False)
persist trainable param layer_norm_25.b_0 : LOD_TENSOR.shape(768,).dtype(FP32).stop_gradient(False)
persist trainable param bert_lm_prediction_head_0.w_0 : LOD_TENSOR.shape(768, 30528).dtype(FP32).stop_gradient(False)
persist trainable param bert_lm_prediction_head_0.b_0 : LOD_TENSOR.shape(30528,).dtype(FP32).stop_gradient(False)
persist trainable param linear_2.w_0 : LOD_TENSOR.shape(768, 2).dtype(FP32).stop_gradient(False)
persist trainable param embedding_2.w_0 : LOD_TENSOR.shape(7633, 768).dtype(FP32).stop_gradient(False)
persist trainable param linear_3.w_0 : LOD_TENSOR.shape(768, 192).dtype(FP32).stop_gradient(False)
persist trainable param linear_4.w_0 : LOD_TENSOR.shape(768, 192).dtype(FP32).stop_gradient(False)
persist trainable param linear_5.w_0 : LOD_TENSOR.shape(768, 192).dtype(FP32).stop_gradient(False)
persist trainable param linear_6.w_0 : LOD_TENSOR.shape(192, 768).dtype(FP32).stop_gradient(False)
persist trainable param linear_7.w_0 : LOD_TENSOR.shape(768, 768).dtype(FP32).stop_gradient(False)
persist trainable param linear_8.w_0 : LOD_TENSOR.shape(768, 768).dtype(FP32).stop_gradient(False)
persist trainable param linear_9.w_0 : LOD_TENSOR.shape(768, 192).dtype(FP32).stop_gradient(False)
persist trainable param linear_10.w_0 : LOD_TENSOR.shape(768, 192).dtype(FP32).stop_gradient(False)
persist trainable param linear_11.w_0 : LOD_TENSOR.shape(768, 192).dtype(FP32).stop_gradient(False)
persist trainable param linear_12.w_0 : LOD_TENSOR.shape(192, 768).dtype(FP32).stop_gradient(False)
persist trainable param linear_13.w_0 : LOD_TENSOR.shape(768, 768).dtype(FP32).stop_gradient(False)
persist trainable param linear_14.w_0 : LOD_TENSOR.shape(768, 768).dtype(FP32).stop_gradient(False)
persist trainable param linear_15.w_0 : LOD_TENSOR.shape(768, 192).dtype(FP32).stop_gradient(False)
persist trainable param linear_16.w_0 : LOD_TENSOR.shape(768, 192).dtype(FP32).stop_gradient(False)
persist trainable param linear_17.w_0 : LOD_TENSOR.shape(768, 192).dtype(FP32).stop_gradient(False)
persist trainable param linear_18.w_0 : LOD_TENSOR.shape(192, 768).dtype(FP32).stop_gradient(False)
persist trainable param linear_19.w_0 : LOD_TENSOR.shape(768, 768).dtype(FP32).stop_gradient(False)
persist trainable param linear_20.w_0 : LOD_TENSOR.shape(768, 768).dtype(FP32).stop_gradient(False)
persist trainable param linear_21.w_0 : LOD_TENSOR.shape(768, 192).dtype(FP32).stop_gradient(False)
persist trainable param linear_22.w_0 : LOD_TENSOR.shape(768, 192).dtype(FP32).stop_gradient(False)
persist trainable param linear_23.w_0 : LOD_TENSOR.shape(768, 192).dtype(FP32).stop_gradient(False)
persist trainable param linear_24.w_0 : LOD_TENSOR.shape(192, 768).dtype(FP32).stop_gradient(False)
persist trainable param linear_25.w_0 : LOD_TENSOR.shape(768, 768).dtype(FP32).stop_gradient(False)
persist trainable param linear_26.w_0 : LOD_TENSOR.shape(768, 768).dtype(FP32).stop_gradient(False)
persist trainable param linear_27.w_0 : LOD_TENSOR.shape(768, 192).dtype(FP32).stop_gradient(False)
persist trainable param linear_28.w_0 : LOD_TENSOR.shape(768, 192).dtype(FP32).stop_gradient(False)
persist trainable param linear_29.w_0 : LOD_TENSOR.shape(768, 192).dtype(FP32).stop_gradient(False)
persist trainable param linear_30.w_0 : LOD_TENSOR.shape(192, 768).dtype(FP32).stop_gradient(False)
persist trainable param linear_31.w_0 : LOD_TENSOR.shape(768, 768).dtype(FP32).stop_gradient(False)
persist trainable param linear_32.w_0 : LOD_TENSOR.shape(768, 768).dtype(FP32).stop_gradient(False)
persist trainable param linear_33.w_0 : LOD_TENSOR.shape(768, 192).dtype(FP32).stop_gradient(False)
persist trainable param linear_34.w_0 : LOD_TENSOR.shape(768, 192).dtype(FP32).stop_gradient(False)
persist trainable param linear_35.w_0 : LOD_TENSOR.shape(768, 192).dtype(FP32).stop_gradient(False)
persist trainable param linear_36.w_0 : LOD_TENSOR.shape(192, 768).dtype(FP32).stop_gradient(False)
persist trainable param linear_37.w_0 : LOD_TENSOR.shape(768, 768).dtype(FP32).stop_gradient(False)
persist trainable param linear_38.w_0 : LOD_TENSOR.shape(768, 768).dtype(FP32).stop_gradient(False)
persist trainable param linear_39.w_0 : LOD_TENSOR.shape(768, 192).dtype(FP32).stop_gradient(False)
persist trainable param linear_40.w_0 : LOD_TENSOR.shape(768, 192).dtype(FP32).stop_gradient(False)
persist trainable param linear_41.w_0 : LOD_TENSOR.shape(768, 192).dtype(FP32).stop_gradient(False)
persist trainable param linear_42.w_0 : LOD_TENSOR.shape(192, 768).dtype(FP32).stop_gradient(False)
persist trainable param linear_43.w_0 : LOD_TENSOR.shape(768, 768).dtype(FP32).stop_gradient(False)
persist trainable param linear_44.w_0 : LOD_TENSOR.shape(768, 768).dtype(FP32).stop_gradient(False)
persist trainable param linear_45.w_0 : LOD_TENSOR.shape(768, 192).dtype(FP32).stop_gradient(False)
persist trainable param linear_46.w_0 : LOD_TENSOR.shape(768, 192).dtype(FP32).stop_gradient(False)
persist trainable param linear_47.w_0 : LOD_TENSOR.shape(768, 192).dtype(FP32).stop_gradient(False)
persist trainable param linear_48.w_0 : LOD_TENSOR.shape(192, 768).dtype(FP32).stop_gradient(False)
persist trainable param linear_49.w_0 : LOD_TENSOR.shape(768, 768).dtype(FP32).stop_gradient(False)
persist trainable param linear_50.w_0 : LOD_TENSOR.shape(768, 768).dtype(FP32).stop_gradient(False)
persist trainable param linear_51.w_0 : LOD_TENSOR.shape(768, 192).dtype(FP32).stop_gradient(False)
persist trainable param linear_52.w_0 : LOD_TENSOR.shape(768, 192).dtype(FP32).stop_gradient(False)
persist trainable param linear_53.w_0 : LOD_TENSOR.shape(768, 192).dtype(FP32).stop_gradient(False)
persist trainable param linear_54.w_0 : LOD_TENSOR.shape(192, 768).dtype(FP32).stop_gradient(False)
persist trainable param linear_55.w_0 : LOD_TENSOR.shape(768, 768).dtype(FP32).stop_gradient(False)
persist trainable param linear_56.w_0 : LOD_TENSOR.shape(768, 768).dtype(FP32).stop_gradient(False)
persist trainable param linear_57.w_0 : LOD_TENSOR.shape(768, 192).dtype(FP32).stop_gradient(False)
persist trainable param linear_58.w_0 : LOD_TENSOR.shape(768, 192).dtype(FP32).stop_gradient(False)
persist trainable param linear_59.w_0 : LOD_TENSOR.shape(768, 192).dtype(FP32).stop_gradient(False)
persist trainable param linear_60.w_0 : LOD_TENSOR.shape(192, 768).dtype(FP32).stop_gradient(False)
persist trainable param linear_61.w_0 : LOD_TENSOR.shape(768, 768).dtype(FP32).stop_gradient(False)
persist trainable param linear_62.w_0 : LOD_TENSOR.shape(768, 768).dtype(FP32).stop_gradient(False)
persist trainable param linear_63.w_0 : LOD_TENSOR.shape(768, 192).dtype(FP32).stop_gradient(False)
persist trainable param linear_64.w_0 : LOD_TENSOR.shape(768, 192).dtype(FP32).stop_gradient(False)
persist trainable param linear_65.w_0 : LOD_TENSOR.shape(768, 192).dtype(FP32).stop_gradient(False)
persist trainable param linear_66.w_0 : LOD_TENSOR.shape(192, 768).dtype(FP32).stop_gradient(False)
persist trainable param linear_67.w_0 : LOD_TENSOR.shape(768, 768).dtype(FP32).stop_gradient(False)
persist trainable param linear_68.w_0 : LOD_TENSOR.shape(768, 768).dtype(FP32).stop_gradient(False)
persist trainable param linear_69.w_0 : LOD_TENSOR.shape(768, 192).dtype(FP32).stop_gradient(False)
persist trainable param linear_70.w_0 : LOD_TENSOR.shape(768, 192).dtype(FP32).stop_gradient(False)
persist trainable param linear_71.w_0 : LOD_TENSOR.shape(768, 192).dtype(FP32).stop_gradient(False)
persist trainable param linear_72.w_0 : LOD_TENSOR.shape(192, 768).dtype(FP32).stop_gradient(False)
persist trainable param linear_73.w_0 : LOD_TENSOR.shape(768, 768).dtype(FP32).stop_gradient(False)
persist trainable param linear_74.w_0 : LOD_TENSOR.shape(768, 768).dtype(FP32).stop_gradient(False)
persist var bert_lm_prediction_head_0.b_0_moment1_0 : LOD_TENSOR.shape(30528,).dtype(FP32).stop_gradient(False)
persist var bert_lm_prediction_head_0.b_0_moment2_0 : LOD_TENSOR.shape(30528,).dtype(FP32).stop_gradient(False)
persist var bert_lm_prediction_head_0.b_0_beta1_pow_acc_0 : LOD_TENSOR.shape(1,).dtype(FP32).stop_gradient(False)
persist var bert_lm_prediction_head_0.b_0_beta2_pow_acc_0 : LOD_TENSOR.shape(1,).dtype(FP32).stop_gradient(False)
persist var bert_lm_prediction_head_0.w_0_moment1_0 : LOD_TENSOR.shape(768, 30528).dtype(FP32).stop_gradient(False)
persist var bert_lm_prediction_head_0.w_0_moment2_0 : LOD_TENSOR.shape(768, 30528).dtype(FP32).stop_gradient(False)
persist var bert_lm_prediction_head_0.w_0_beta1_pow_acc_0 : LOD_TENSOR.shape(1,).dtype(FP32).stop_gradient(False)
persist var bert_lm_prediction_head_0.w_0_beta2_pow_acc_0 : LOD_TENSOR.shape(1,).dtype(FP32).stop_gradient(False)
persist var embedding_0.w_0_moment1_0 : LOD_TENSOR.shape(512, 768).dtype(FP32).stop_gradient(False)
persist var embedding_0.w_0_moment2_0 : LOD_TENSOR.shape(512, 768).dtype(FP32).stop_gradient(False)
persist var embedding_0.w_0_beta1_pow_acc_0 : LOD_TENSOR.shape(1,).dtype(FP32).stop_gradient(False)
persist var embedding_0.w_0_beta2_pow_acc_0 : LOD_TENSOR.shape(1,).dtype(FP32).stop_gradient(False)
persist var embedding_1.w_0_moment1_0 : LOD_TENSOR.shape(2, 768).dtype(FP32).stop_gradient(False)
persist var embedding_1.w_0_moment2_0 : LOD_TENSOR.shape(2, 768).dtype(FP32).stop_gradient(False)
persist var embedding_1.w_0_beta1_pow_acc_0 : LOD_TENSOR.shape(1,).dtype(FP32).stop_gradient(False)
persist var embedding_1.w_0_beta2_pow_acc_0 : LOD_TENSOR.shape(1,).dtype(FP32).stop_gradient(False)
persist var layer_norm_0.b_0_moment1_0 : LOD_TENSOR.shape(768,).dtype(FP32).stop_gradient(False)
persist var layer_norm_0.b_0_moment2_0 : LOD_TENSOR.shape(768,).dtype(FP32).stop_gradient(False)
persist var layer_norm_0.b_0_beta1_pow_acc_0 : LOD_TENSOR.shape(1,).dtype(FP32).stop_gradient(False)
persist var layer_norm_0.b_0_beta2_pow_acc_0 : LOD_TENSOR.shape(1,).dtype(FP32).stop_gradient(False)
persist var layer_norm_0.w_0_moment1_0 : LOD_TENSOR.shape(768,).dtype(FP32).stop_gradient(False)
persist var layer_norm_0.w_0_moment2_0 : LOD_TENSOR.shape(768,).dtype(FP32).stop_gradient(False)
persist var layer_norm_0.w_0_beta1_pow_acc_0 : LOD_TENSOR.shape(1,).dtype(FP32).stop_gradient(False)
persist var layer_norm_0.w_0_beta2_pow_acc_0 : LOD_TENSOR.shape(1,).dtype(FP32).stop_gradient(False)
persist var linear_0.b_0_moment1_0 : LOD_TENSOR.shape(768,).dtype(FP32).stop_gradient(False)
persist var linear_0.b_0_moment2_0 : LOD_TENSOR.shape(768,).dtype(FP32).stop_gradient(False)
persist var linear_0.b_0_beta1_pow_acc_0 : LOD_TENSOR.shape(1,).dtype(FP32).stop_gradient(False)
persist var linear_0.b_0_beta2_pow_acc_0 : LOD_TENSOR.shape(1,).dtype(FP32).stop_gradient(False)
persist var linear_0.w_0_moment1_0 : LOD_TENSOR.shape(768, 768).dtype(FP32).stop_gradient(False)
persist var linear_0.w_0_moment2_0 : LOD_TENSOR.shape(768, 768).dtype(FP32).stop_gradient(False)
persist var linear_0.w_0_beta1_pow_acc_0 : LOD_TENSOR.shape(1,).dtype(FP32).stop_gradient(False)
persist var linear_0.w_0_beta2_pow_acc_0 : LOD_TENSOR.shape(1,).dtype(FP32).stop_gradient(False)
persist var linear_1.w_0_moment1_0 : LOD_TENSOR.shape(768, 768).dtype(FP32).stop_gradient(False)
persist var linear_1.w_0_moment2_0 : LOD_TENSOR.shape(768, 768).dtype(FP32).stop_gradient(False)
persist var linear_1.w_0_beta1_pow_acc_0 : LOD_TENSOR.shape(1,).dtype(FP32).stop_gradient(False)
persist var linear_1.w_0_beta2_pow_acc_0 : LOD_TENSOR.shape(1,).dtype(FP32).stop_gradient(False)
persist var linear_2.w_0_moment1_0 : LOD_TENSOR.shape(768, 2).dtype(FP32).stop_gradient(False)
persist var linear_2.w_0_moment2_0 : LOD_TENSOR.shape(768, 2).dtype(FP32).stop_gradient(False)
persist var linear_2.w_0_beta1_pow_acc_0 : LOD_TENSOR.shape(1,).dtype(FP32).stop_gradient(False)
persist var linear_2.w_0_beta2_pow_acc_0 : LOD_TENSOR.shape(1,).dtype(FP32).stop_gradient(False)
persist var learning_rate_0 : LOD_TENSOR.shape(1,).dtype(FP32).stop_gradient(False)
persist var NCCLID : RAW)
{Out=['learning_rate_0']} = fill_constant(inputs={}, dtype = 5, force_cpu = False, op_device = , op_namescope = /, op_role = 0, op_role_var = [], shape = [1], str_value = , value = 0.0)
{Out=['linear_2.w_0_beta2_pow_acc_0']} = fill_constant(inputs={}, dtype = 5, force_cpu = False, op_device = cpu, op_namescope = /, op_role = 0, op_role_var = [], shape = [1], str_value = , value = 0.9990000128746033)
{Out=['linear_2.w_0_beta1_pow_acc_0']} = fill_constant(inputs={}, dtype = 5, force_cpu = False, op_device = cpu, op_namescope = /, op_role = 0, op_role_var = [], shape = [1], str_value = , value = 0.8999999761581421)
{Out=['linear_2.w_0_moment2_0']} = fill_constant(inputs={}, dtype = 5, force_cpu = False, op_device = , op_namescope = /, op_role = 0, op_role_var = [], shape = [768, 2], str_value = , value = 0.0)
{Out=['linear_2.w_0_moment1_0']} = fill_constant(inputs={}, dtype = 5, force_cpu = False, op_device = , op_namescope = /, op_role = 0, op_role_var = [], shape = [768, 2], str_value = , value = 0.0)
{Out=['linear_1.w_0_beta2_pow_acc_0']} = fill_constant(inputs={}, dtype = 5, force_cpu = False, op_device = cpu, op_namescope = /, op_role = 0, op_role_var = [], shape = [1], str_value = , value = 0.9990000128746033)
{Out=['linear_1.w_0_beta1_pow_acc_0']} = fill_constant(inputs={}, dtype = 5, force_cpu = False, op_device = cpu, op_namescope = /, op_role = 0, op_role_var = [], shape = [1], str_value = , value = 0.8999999761581421)
{Out=['linear_1.w_0_moment2_0']} = fill_constant(inputs={}, dtype = 5, force_cpu = False, op_device = , op_namescope = /, op_role = 0, op_role_var = [], shape = [768, 768], str_value = , value = 0.0)
{Out=['linear_1.w_0_moment1_0']} = fill_constant(inputs={}, dtype = 5, force_cpu = False, op_device = , op_namescope = /, op_role = 0, op_role_var = [], shape = [768, 768], str_value = , value = 0.0)
{Out=['linear_0.w_0_beta2_pow_acc_0']} = fill_constant(inputs={}, dtype = 5, force_cpu = False, op_device = cpu, op_namescope = /, op_role = 0, op_role_var = [], shape = [1], str_value = , value = 0.9990000128746033)
{Out=['linear_0.w_0_beta1_pow_acc_0']} = fill_constant(inputs={}, dtype = 5, force_cpu = False, op_device = cpu, op_namescope = /, op_role = 0, op_role_var = [], shape = [1], str_value = , value = 0.8999999761581421)
{Out=['linear_0.w_0_moment2_0']} = fill_constant(inputs={}, dtype = 5, force_cpu = False, op_device = , op_namescope = /, op_role = 0, op_role_var = [], shape = [768, 768], str_value = , value = 0.0)
{Out=['linear_0.w_0_moment1_0']} = fill_constant(inputs={}, dtype = 5, force_cpu = False, op_device = , op_namescope = /, op_role = 0, op_role_var = [], shape = [768, 768], str_value = , value = 0.0)
{Out=['linear_0.b_0_beta2_pow_acc_0']} = fill_constant(inputs={}, dtype = 5, force_cpu = False, op_device = cpu, op_namescope = /, op_role = 0, op_role_var = [], shape = [1], str_value = , value = 0.9990000128746033)
{Out=['linear_0.b_0_beta1_pow_acc_0']} = fill_constant(inputs={}, dtype = 5, force_cpu = False, op_device = cpu, op_namescope = /, op_role = 0, op_role_var = [], shape = [1], str_value = , value = 0.8999999761581421)
{Out=['linear_0.b_0_moment2_0']} = fill_constant(inputs={}, dtype = 5, force_cpu = False, op_device = , op_namescope = /, op_role = 0, op_role_var = [], shape = [768], str_value = , value = 0.0)
{Out=['linear_0.b_0_moment1_0']} = fill_constant(inputs={}, dtype = 5, force_cpu = False, op_device = , op_namescope = /, op_role = 0, op_role_var = [], shape = [768], str_value = , value = 0.0)
{Out=['layer_norm_0.w_0_beta2_pow_acc_0']} = fill_constant(inputs={}, dtype = 5, force_cpu = False, op_device = cpu, op_namescope = /, op_role = 0, op_role_var = [], shape = [1], str_value = , value = 0.9990000128746033)
{Out=['layer_norm_0.w_0_beta1_pow_acc_0']} = fill_constant(inputs={}, dtype = 5, force_cpu = False, op_device = cpu, op_namescope = /, op_role = 0, op_role_var = [], shape = [1], str_value = , value = 0.8999999761581421)
{Out=['layer_norm_0.w_0_moment2_0']} = fill_constant(inputs={}, dtype = 5, force_cpu = False, op_device = , op_namescope = /, op_role = 0, op_role_var = [], shape = [768], str_value = , value = 0.0)
{Out=['layer_norm_0.w_0_moment1_0']} = fill_constant(inputs={}, dtype = 5, force_cpu = False, op_device = , op_namescope = /, op_role = 0, op_role_var = [], shape = [768], str_value = , value = 0.0)
{Out=['layer_norm_0.b_0_beta2_pow_acc_0']} = fill_constant(inputs={}, dtype = 5, force_cpu = False, op_device = cpu, op_namescope = /, op_role = 0, op_role_var = [], shape = [1], str_value = , value = 0.9990000128746033)
{Out=['layer_norm_0.b_0_beta1_pow_acc_0']} = fill_constant(inputs={}, dtype = 5, force_cpu = False, op_device = cpu, op_namescope = /, op_role = 0, op_role_var = [], shape = [1], str_value = , value = 0.8999999761581421)
{Out=['layer_norm_0.b_0_moment2_0']} = fill_constant(inputs={}, dtype = 5, force_cpu = False, op_device = , op_namescope = /, op_role = 0, op_role_var = [], shape = [768], str_value = , value = 0.0)
{Out=['layer_norm_0.b_0_moment1_0']} = fill_constant(inputs={}, dtype = 5, force_cpu = False, op_device = , op_namescope = /, op_role = 0, op_role_var = [], shape = [768], str_value = , value = 0.0)
{Out=['embedding_1.w_0_beta2_pow_acc_0']} = fill_constant(inputs={}, dtype = 5, force_cpu = False, op_device = cpu, op_namescope = /, op_role = 0, op_role_var = [], shape = [1], str_value = , value = 0.9990000128746033)
{Out=['embedding_1.w_0_beta1_pow_acc_0']} = fill_constant(inputs={}, dtype = 5, force_cpu = False, op_device = cpu, op_namescope = /, op_role = 0, op_role_var = [], shape = [1], str_value = , value = 0.8999999761581421)
{Out=['embedding_1.w_0_moment2_0']} = fill_constant(inputs={}, dtype = 5, force_cpu = False, op_device = , op_namescope = /, op_role = 0, op_role_var = [], shape = [2, 768], str_value = , value = 0.0)
{Out=['embedding_1.w_0_moment1_0']} = fill_constant(inputs={}, dtype = 5, force_cpu = False, op_device = , op_namescope = /, op_role = 0, op_role_var = [], shape = [2, 768], str_value = , value = 0.0)
{Out=['embedding_0.w_0_beta2_pow_acc_0']} = fill_constant(inputs={}, dtype = 5, force_cpu = False, op_device = cpu, op_namescope = /, op_role = 0, op_role_var = [], shape = [1], str_value = , value = 0.9990000128746033)
{Out=['embedding_0.w_0_beta1_pow_acc_0']} = fill_constant(inputs={}, dtype = 5, force_cpu = False, op_device = cpu, op_namescope = /, op_role = 0, op_role_var = [], shape = [1], str_value = , value = 0.8999999761581421)
{Out=['embedding_0.w_0_moment2_0']} = fill_constant(inputs={}, dtype = 5, force_cpu = False, op_device = , op_namescope = /, op_role = 0, op_role_var = [], shape = [512, 768], str_value = , value = 0.0)
{Out=['embedding_0.w_0_moment1_0']} = fill_constant(inputs={}, dtype = 5, force_cpu = False, op_device = , op_namescope = /, op_role = 0, op_role_var = [], shape = [512, 768], str_value = , value = 0.0)
{Out=['bert_lm_prediction_head_0.w_0_beta2_pow_acc_0']} = fill_constant(inputs={}, dtype = 5, force_cpu = False, op_device = cpu, op_namescope = /, op_role = 0, op_role_var = [], shape = [1], str_value = , value = 0.9990000128746033)
{Out=['bert_lm_prediction_head_0.w_0_beta1_pow_acc_0']} = fill_constant(inputs={}, dtype = 5, force_cpu = False, op_device = cpu, op_namescope = /, op_role = 0, op_role_var = [], shape = [1], str_value = , value = 0.8999999761581421)
{Out=['bert_lm_prediction_head_0.w_0_moment2_0']} = fill_constant(inputs={}, dtype = 5, force_cpu = False, op_device = , op_namescope = /, op_role = 0, op_role_var = [], shape = [768, 30528], str_value = , value = 0.0)
{Out=['bert_lm_prediction_head_0.w_0_moment1_0']} = fill_constant(inputs={}, dtype = 5, force_cpu = False, op_device = , op_namescope = /, op_role = 0, op_role_var = [], shape = [768, 30528], str_value = , value = 0.0)
{Out=['bert_lm_prediction_head_0.b_0_beta2_pow_acc_0']} = fill_constant(inputs={}, dtype = 5, force_cpu = False, op_device = cpu, op_namescope = /, op_role = 0, op_role_var = [], shape = [1], str_value = , value = 0.9990000128746033)
{Out=['bert_lm_prediction_head_0.b_0_beta1_pow_acc_0']} = fill_constant(inputs={}, dtype = 5, force_cpu = False, op_device = cpu, op_namescope = /, op_role = 0, op_role_var = [], shape = [1], str_value = , value = 0.8999999761581421)
{Out=['bert_lm_prediction_head_0.b_0_moment2_0']} = fill_constant(inputs={}, dtype = 5, force_cpu = False, op_device = , op_namescope = /, op_role = 0, op_role_var = [], shape = [30528], str_value = , value = 0.0)
{Out=['bert_lm_prediction_head_0.b_0_moment1_0']} = fill_constant(inputs={}, dtype = 5, force_cpu = False, op_device = , op_namescope = /, op_role = 0, op_role_var = [], shape = [30528], str_value = , value = 0.0)
{Out=['linear_74.w_0']} = fill_constant(inputs={}, dtype = 5, force_cpu = False, op_device = , op_namescope = /, op_role = 0, op_role_var = [], shape = [768, 768], str_value = , value = 0.009999999776482582)
{Out=['linear_73.w_0']} = fill_constant(inputs={}, dtype = 5, force_cpu = False, op_device = , op_namescope = /, op_role = 0, op_role_var = [], shape = [768, 768], str_value = , value = 0.009999999776482582)
{Out=['linear_72.w_0']} = fill_constant(inputs={}, dtype = 5, force_cpu = False, op_device = , op_namescope = /, op_role = 0, op_role_var = [], shape = [192, 768], str_value = , value = 0.009999999776482582)
{Out=['linear_71.w_0']} = fill_constant(inputs={}, dtype = 5, force_cpu = False, op_device = , op_namescope = /, op_role = 0, op_role_var = [], shape = [768, 192], str_value = , value = 0.009999999776482582)
{Out=['linear_70.w_0']} = fill_constant(inputs={}, dtype = 5, force_cpu = False, op_device = , op_namescope = /, op_role = 0, op_role_var = [], shape = [768, 192], str_value = , value = 0.009999999776482582)
{Out=['linear_69.w_0']} = fill_constant(inputs={}, dtype = 5, force_cpu = False, op_device = , op_namescope = /, op_role = 0, op_role_var = [], shape = [768, 192], str_value = , value = 0.009999999776482582)
{Out=['linear_68.w_0']} = fill_constant(inputs={}, dtype = 5, force_cpu = False, op_device = , op_namescope = /, op_role = 0, op_role_var = [], shape = [768, 768], str_value = , value = 0.009999999776482582)
{Out=['linear_67.w_0']} = fill_constant(inputs={}, dtype = 5, force_cpu = False, op_device = , op_namescope = /, op_role = 0, op_role_var = [], shape = [768, 768], str_value = , value = 0.009999999776482582)
{Out=['linear_66.w_0']} = fill_constant(inputs={}, dtype = 5, force_cpu = False, op_device = , op_namescope = /, op_role = 0, op_role_var = [], shape = [192, 768], str_value = , value = 0.009999999776482582)
{Out=['linear_65.w_0']} = fill_constant(inputs={}, dtype = 5, force_cpu = False, op_device = , op_namescope = /, op_role = 0, op_role_var = [], shape = [768, 192], str_value = , value = 0.009999999776482582)
{Out=['linear_64.w_0']} = fill_constant(inputs={}, dtype = 5, force_cpu = False, op_device = , op_namescope = /, op_role = 0, op_role_var = [], shape = [768, 192], str_value = , value = 0.009999999776482582)
{Out=['linear_63.w_0']} = fill_constant(inputs={}, dtype = 5, force_cpu = False, op_device = , op_namescope = /, op_role = 0, op_role_var = [], shape = [768, 192], str_value = , value = 0.009999999776482582)
{Out=['linear_62.w_0']} = fill_constant(inputs={}, dtype = 5, force_cpu = False, op_device = , op_namescope = /, op_role = 0, op_role_var = [], shape = [768, 768], str_value = , value = 0.009999999776482582)
{Out=['linear_61.w_0']} = fill_constant(inputs={}, dtype = 5, force_cpu = False, op_device = , op_namescope = /, op_role = 0, op_role_var = [], shape = [768, 768], str_value = , value = 0.009999999776482582)
{Out=['linear_60.w_0']} = fill_constant(inputs={}, dtype = 5, force_cpu = False, op_device = , op_namescope = /, op_role = 0, op_role_var = [], shape = [192, 768], str_value = , value = 0.009999999776482582)
{Out=['linear_59.w_0']} = fill_constant(inputs={}, dtype = 5, force_cpu = False, op_device = , op_namescope = /, op_role = 0, op_role_var = [], shape = [768, 192], str_value = , value = 0.009999999776482582)
{Out=['linear_58.w_0']} = fill_constant(inputs={}, dtype = 5, force_cpu = False, op_device = , op_namescope = /, op_role = 0, op_role_var = [], shape = [768, 192], str_value = , value = 0.009999999776482582)
{Out=['linear_57.w_0']} = fill_constant(inputs={}, dtype = 5, force_cpu = False, op_device = , op_namescope = /, op_role = 0, op_role_var = [], shape = [768, 192], str_value = , value = 0.009999999776482582)
{Out=['linear_56.w_0']} = fill_constant(inputs={}, dtype = 5, force_cpu = False, op_device = , op_namescope = /, op_role = 0, op_role_var = [], shape = [768, 768], str_value = , value = 0.009999999776482582)
{Out=['linear_55.w_0']} = fill_constant(inputs={}, dtype = 5, force_cpu = False, op_device = , op_namescope = /, op_role = 0, op_role_var = [], shape = [768, 768], str_value = , value = 0.009999999776482582)
{Out=['linear_54.w_0']} = fill_constant(inputs={}, dtype = 5, force_cpu = False, op_device = , op_namescope = /, op_role = 0, op_role_var = [], shape = [192, 768], str_value = , value = 0.009999999776482582)
{Out=['linear_53.w_0']} = fill_constant(inputs={}, dtype = 5, force_cpu = False, op_device = , op_namescope = /, op_role = 0, op_role_var = [], shape = [768, 192], str_value = , value = 0.009999999776482582)
{Out=['linear_52.w_0']} = fill_constant(inputs={}, dtype = 5, force_cpu = False, op_device = , op_namescope = /, op_role = 0, op_role_var = [], shape = [768, 192], str_value = , value = 0.009999999776482582)
{Out=['linear_51.w_0']} = fill_constant(inputs={}, dtype = 5, force_cpu = False, op_device = , op_namescope = /, op_role = 0, op_role_var = [], shape = [768, 192], str_value = , value = 0.009999999776482582)
{Out=['linear_50.w_0']} = fill_constant(inputs={}, dtype = 5, force_cpu = False, op_device = , op_namescope = /, op_role = 0, op_role_var = [], shape = [768, 768], str_value = , value = 0.009999999776482582)
{Out=['linear_49.w_0']} = fill_constant(inputs={}, dtype = 5, force_cpu = False, op_device = , op_namescope = /, op_role = 0, op_role_var = [], shape = [768, 768], str_value = , value = 0.009999999776482582)
{Out=['linear_48.w_0']} = fill_constant(inputs={}, dtype = 5, force_cpu = False, op_device = , op_namescope = /, op_role = 0, op_role_var = [], shape = [192, 768], str_value = , value = 0.009999999776482582)
{Out=['linear_47.w_0']} = fill_constant(inputs={}, dtype = 5, force_cpu = False, op_device = , op_namescope = /, op_role = 0, op_role_var = [], shape = [768, 192], str_value = , value = 0.009999999776482582)
{Out=['linear_46.w_0']} = fill_constant(inputs={}, dtype = 5, force_cpu = False, op_device = , op_namescope = /, op_role = 0, op_role_var = [], shape = [768, 192], str_value = , value = 0.009999999776482582)
{Out=['linear_45.w_0']} = fill_constant(inputs={}, dtype = 5, force_cpu = False, op_device = , op_namescope = /, op_role = 0, op_role_var = [], shape = [768, 192], str_value = , value = 0.009999999776482582)
{Out=['linear_44.w_0']} = fill_constant(inputs={}, dtype = 5, force_cpu = False, op_device = , op_namescope = /, op_role = 0, op_role_var = [], shape = [768, 768], str_value = , value = 0.009999999776482582)
{Out=['linear_43.w_0']} = fill_constant(inputs={}, dtype = 5, force_cpu = False, op_device = , op_namescope = /, op_role = 0, op_role_var = [], shape = [768, 768], str_value = , value = 0.009999999776482582)
{Out=['linear_42.w_0']} = fill_constant(inputs={}, dtype = 5, force_cpu = False, op_device = , op_namescope = /, op_role = 0, op_role_var = [], shape = [192, 768], str_value = , value = 0.009999999776482582)
{Out=['linear_41.w_0']} = fill_constant(inputs={}, dtype = 5, force_cpu = False, op_device = , op_namescope = /, op_role = 0, op_role_var = [], shape = [768, 192], str_value = , value = 0.009999999776482582)
{Out=['linear_40.w_0']} = fill_constant(inputs={}, dtype = 5, force_cpu = False, op_device = , op_namescope = /, op_role = 0, op_role_var = [], shape = [768, 192], str_value = , value = 0.009999999776482582)
{Out=['linear_39.w_0']} = fill_constant(inputs={}, dtype = 5, force_cpu = False, op_device = , op_namescope = /, op_role = 0, op_role_var = [], shape = [768, 192], str_value = , value = 0.009999999776482582)
{Out=['linear_38.w_0']} = fill_constant(inputs={}, dtype = 5, force_cpu = False, op_device = , op_namescope = /, op_role = 0, op_role_var = [], shape = [768, 768], str_value = , value = 0.009999999776482582)
{Out=['linear_37.w_0']} = fill_constant(inputs={}, dtype = 5, force_cpu = False, op_device = , op_namescope = /, op_role = 0, op_role_var = [], shape = [768, 768], str_value = , value = 0.009999999776482582)
{Out=['linear_36.w_0']} = fill_constant(inputs={}, dtype = 5, force_cpu = False, op_device = , op_namescope = /, op_role = 0, op_role_var = [], shape = [192, 768], str_value = , value = 0.009999999776482582)
{Out=['linear_35.w_0']} = fill_constant(inputs={}, dtype = 5, force_cpu = False, op_device = , op_namescope = /, op_role = 0, op_role_var = [], shape = [768, 192], str_value = , value = 0.009999999776482582)
{Out=['linear_34.w_0']} = fill_constant(inputs={}, dtype = 5, force_cpu = False, op_device = , op_namescope = /, op_role = 0, op_role_var = [], shape = [768, 192], str_value = , value = 0.009999999776482582)
{Out=['linear_33.w_0']} = fill_constant(inputs={}, dtype = 5, force_cpu = False, op_device = , op_namescope = /, op_role = 0, op_role_var = [], shape = [768, 192], str_value = , value = 0.009999999776482582)
{Out=['linear_32.w_0']} = fill_constant(inputs={}, dtype = 5, force_cpu = False, op_device = , op_namescope = /, op_role = 0, op_role_var = [], shape = [768, 768], str_value = , value = 0.009999999776482582)
{Out=['linear_31.w_0']} = fill_constant(inputs={}, dtype = 5, force_cpu = False, op_device = , op_namescope = /, op_role = 0, op_role_var = [], shape = [768, 768], str_value = , value = 0.009999999776482582)
{Out=['linear_30.w_0']} = fill_constant(inputs={}, dtype = 5, force_cpu = False, op_device = , op_namescope = /, op_role = 0, op_role_var = [], shape = [192, 768], str_value = , value = 0.009999999776482582)
{Out=['linear_29.w_0']} = fill_constant(inputs={}, dtype = 5, force_cpu = False, op_device = , op_namescope = /, op_role = 0, op_role_var = [], shape = [768, 192], str_value = , value = 0.009999999776482582)
{Out=['linear_28.w_0']} = fill_constant(inputs={}, dtype = 5, force_cpu = False, op_device = , op_namescope = /, op_role = 0, op_role_var = [], shape = [768, 192], str_value = , value = 0.009999999776482582)
{Out=['linear_27.w_0']} = fill_constant(inputs={}, dtype = 5, force_cpu = False, op_device = , op_namescope = /, op_role = 0, op_role_var = [], shape = [768, 192], str_value = , value = 0.009999999776482582)
{Out=['linear_26.w_0']} = fill_constant(inputs={}, dtype = 5, force_cpu = False, op_device = , op_namescope = /, op_role = 0, op_role_var = [], shape = [768, 768], str_value = , value = 0.009999999776482582)
{Out=['linear_25.w_0']} = fill_constant(inputs={}, dtype = 5, force_cpu = False, op_device = , op_namescope = /, op_role = 0, op_role_var = [], shape = [768, 768], str_value = , value = 0.009999999776482582)
{Out=['linear_24.w_0']} = fill_constant(inputs={}, dtype = 5, force_cpu = False, op_device = , op_namescope = /, op_role = 0, op_role_var = [], shape = [192, 768], str_value = , value = 0.009999999776482582)
{Out=['linear_23.w_0']} = fill_constant(inputs={}, dtype = 5, force_cpu = False, op_device = , op_namescope = /, op_role = 0, op_role_var = [], shape = [768, 192], str_value = , value = 0.009999999776482582)
{Out=['linear_22.w_0']} = fill_constant(inputs={}, dtype = 5, force_cpu = False, op_device = , op_namescope = /, op_role = 0, op_role_var = [], shape = [768, 192], str_value = , value = 0.009999999776482582)
{Out=['linear_21.w_0']} = fill_constant(inputs={}, dtype = 5, force_cpu = False, op_device = , op_namescope = /, op_role = 0, op_role_var = [], shape = [768, 192], str_value = , value = 0.009999999776482582)
{Out=['linear_20.w_0']} = fill_constant(inputs={}, dtype = 5, force_cpu = False, op_device = , op_namescope = /, op_role = 0, op_role_var = [], shape = [768, 768], str_value = , value = 0.009999999776482582)
{Out=['linear_19.w_0']} = fill_constant(inputs={}, dtype = 5, force_cpu = False, op_device = , op_namescope = /, op_role = 0, op_role_var = [], shape = [768, 768], str_value = , value = 0.009999999776482582)
{Out=['linear_18.w_0']} = fill_constant(inputs={}, dtype = 5, force_cpu = False, op_device = , op_namescope = /, op_role = 0, op_role_var = [], shape = [192, 768], str_value = , value = 0.009999999776482582)
{Out=['linear_17.w_0']} = fill_constant(inputs={}, dtype = 5, force_cpu = False, op_device = , op_namescope = /, op_role = 0, op_role_var = [], shape = [768, 192], str_value = , value = 0.009999999776482582)
{Out=['linear_16.w_0']} = fill_constant(inputs={}, dtype = 5, force_cpu = False, op_device = , op_namescope = /, op_role = 0, op_role_var = [], shape = [768, 192], str_value = , value = 0.009999999776482582)
{Out=['linear_15.w_0']} = fill_constant(inputs={}, dtype = 5, force_cpu = False, op_device = , op_namescope = /, op_role = 0, op_role_var = [], shape = [768, 192], str_value = , value = 0.009999999776482582)
{Out=['linear_14.w_0']} = fill_constant(inputs={}, dtype = 5, force_cpu = False, op_device = , op_namescope = /, op_role = 0, op_role_var = [], shape = [768, 768], str_value = , value = 0.009999999776482582)
{Out=['linear_13.w_0']} = fill_constant(inputs={}, dtype = 5, force_cpu = False, op_device = , op_namescope = /, op_role = 0, op_role_var = [], shape = [768, 768], str_value = , value = 0.009999999776482582)
{Out=['linear_12.w_0']} = fill_constant(inputs={}, dtype = 5, force_cpu = False, op_device = , op_namescope = /, op_role = 0, op_role_var = [], shape = [192, 768], str_value = , value = 0.009999999776482582)
{Out=['linear_11.w_0']} = fill_constant(inputs={}, dtype = 5, force_cpu = False, op_device = , op_namescope = /, op_role = 0, op_role_var = [], shape = [768, 192], str_value = , value = 0.009999999776482582)
{Out=['linear_10.w_0']} = fill_constant(inputs={}, dtype = 5, force_cpu = False, op_device = , op_namescope = /, op_role = 0, op_role_var = [], shape = [768, 192], str_value = , value = 0.009999999776482582)
{Out=['linear_9.w_0']} = fill_constant(inputs={}, dtype = 5, force_cpu = False, op_device = , op_namescope = /, op_role = 0, op_role_var = [], shape = [768, 192], str_value = , value = 0.009999999776482582)
{Out=['linear_8.w_0']} = fill_constant(inputs={}, dtype = 5, force_cpu = False, op_device = , op_namescope = /, op_role = 0, op_role_var = [], shape = [768, 768], str_value = , value = 0.009999999776482582)
{Out=['linear_7.w_0']} = fill_constant(inputs={}, dtype = 5, force_cpu = False, op_device = , op_namescope = /, op_role = 0, op_role_var = [], shape = [768, 768], str_value = , value = 0.009999999776482582)
{Out=['linear_6.w_0']} = fill_constant(inputs={}, dtype = 5, force_cpu = False, op_device = , op_namescope = /, op_role = 0, op_role_var = [], shape = [192, 768], str_value = , value = 0.009999999776482582)
{Out=['linear_5.w_0']} = fill_constant(inputs={}, dtype = 5, force_cpu = False, op_device = , op_namescope = /, op_role = 0, op_role_var = [], shape = [768, 192], str_value = , value = 0.009999999776482582)
{Out=['linear_4.w_0']} = fill_constant(inputs={}, dtype = 5, force_cpu = False, op_device = , op_namescope = /, op_role = 0, op_role_var = [], shape = [768, 192], str_value = , value = 0.009999999776482582)
{Out=['linear_3.w_0']} = fill_constant(inputs={}, dtype = 5, force_cpu = False, op_device = , op_namescope = /, op_role = 0, op_role_var = [], shape = [768, 192], str_value = , value = 0.009999999776482582)
{Out=['embedding_2.w_0']} = fill_constant(inputs={}, dtype = 5, force_cpu = False, op_device = , op_namescope = /, op_role = 0, op_role_var = [], shape = [7633, 768], str_value = , value = 9.999999974752427e-07)
{Out=['linear_2.w_0']} = fill_constant(inputs={}, dtype = 5, force_cpu = False, op_device = , op_namescope = /, op_role = 0, op_role_var = [], shape = [768, 2], str_value = , value = 9.999999974752427e-07)
{Out=['bert_lm_prediction_head_0.b_0']} = fill_constant(inputs={}, dtype = 5, force_cpu = False, op_device = , op_namescope = /, op_role = 0, op_role_var = [], shape = [30528], str_value = , value = 0.0)
{Out=['bert_lm_prediction_head_0.w_0']} = uniform_random(inputs={ShapeTensor=[], ShapeTensorList=[]}, diag_num = 0, diag_step = 0, diag_val = 1.0, dtype = 5, max = 0.013846219517290592, min = -0.013846219517290592, op_device = , op_namescope = /, op_role = 0, op_role_var = [], seed = 0, shape = [768, 30528])
{Out=['layer_norm_25.b_0']} = fill_constant(inputs={}, dtype = 5, force_cpu = False, op_device = , op_namescope = /, op_role = 0, op_role_var = [], shape = [768], str_value = , value = 0.0)
{Out=['layer_norm_25.w_0']} = fill_constant(inputs={}, dtype = 5, force_cpu = False, op_device = , op_namescope = /, op_role = 0, op_role_var = [], shape = [768], str_value = , value = 1.0)
{Out=['linear_1.w_0']} = fill_constant(inputs={}, dtype = 5, force_cpu = False, op_device = , op_namescope = /, op_role = 0, op_role_var = [], shape = [768, 768], str_value = , value = 9.999999974752427e-07)
{Out=['linear_0.b_0']} = fill_constant(inputs={}, dtype = 5, force_cpu = False, op_device = , op_namescope = /, op_role = 0, op_role_var = [], shape = [768], str_value = , value = 0.0)
{Out=['linear_0.w_0']} = fill_constant(inputs={}, dtype = 5, force_cpu = False, op_device = , op_namescope = /, op_role = 0, op_role_var = [], shape = [768, 768], str_value = , value = 0.009999999776482582)
{Out=['layer_norm_0.b_0']} = fill_constant(inputs={}, dtype = 5, force_cpu = False, op_device = , op_namescope = /, op_role = 0, op_role_var = [], shape = [768], str_value = , value = 0.0)
{Out=['layer_norm_0.w_0']} = fill_constant(inputs={}, dtype = 5, force_cpu = False, op_device = , op_namescope = /, op_role = 0, op_role_var = [], shape = [768], str_value = , value = 1.0)
{Out=['embedding_1.w_0']} = fill_constant(inputs={}, dtype = 5, force_cpu = False, op_device = , op_namescope = /, op_role = 0, op_role_var = [], shape = [2, 768], str_value = , value = 9.999999974752427e-07)
{Out=['embedding_0.w_0']} = fill_constant(inputs={}, dtype = 5, force_cpu = False, op_device = , op_namescope = /, op_role = 0, op_role_var = [], shape = [512, 768], str_value = , value = 9.999999974752427e-07)
{NCCLID=['NCCLID']} = gen_nccl_id(inputs={}, hierarchical_allreduce_inter_nranks = -1, nccl_comm_num = 1, op_device = , op_namescope = /, op_role = 0, op_role_var = [], trainer_id = 0, trainers = ['127.0.0.1:24234', '127.0.0.1:26513', '127.0.0.1:59146', '127.0.0.1:26501'], use_hierarchical_allreduce = False)
c_comm_init(inputs={X=['NCCLID']}, device_id = -1, nranks = 4, op_device = , op_namescope = /, op_role = 0, op_role_var = [], rank = 0, ring_id = 0)
}