-
Notifications
You must be signed in to change notification settings - Fork 1
/
Model.py
467 lines (371 loc) · 16.1 KB
/
Model.py
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
297
298
299
300
301
302
303
304
305
306
307
308
309
310
311
312
313
314
315
316
317
318
319
320
321
322
323
324
325
326
327
328
329
330
331
332
333
334
335
336
337
338
339
340
341
342
343
344
345
346
347
348
349
350
351
352
353
354
355
356
357
358
359
360
361
362
363
364
365
366
367
368
369
370
371
372
373
374
375
376
377
378
379
380
381
382
383
384
385
386
387
388
389
390
391
392
393
394
395
396
397
398
399
400
401
402
403
404
405
406
407
408
409
410
411
412
413
414
415
416
417
418
419
420
421
422
423
424
425
426
427
428
429
430
431
432
433
434
435
436
437
438
439
440
441
442
443
444
445
446
447
448
449
450
451
452
453
454
455
456
457
458
459
460
461
462
463
464
465
466
467
# encoding: utf-8
import torch
import numpy as np
import torch.nn as nn
from torch.utils.data import Dataset, DataLoader
class MyDataset(Dataset):
def __init__(self, file_name):
super(MyDataset, self).__init__()
self.trainDataset = np.load(file_name, allow_pickle=True)
def __len__(self):
return len(self.trainDataset)
def __getitem__(self, idx):
query = torch.Tensor([self.trainDataset[idx, 0]])
target = self.trainDataset[idx, 1]
return query, target
class AudioQuery(nn.Module):
def __init__(self, out_dim):
super(AudioQuery, self).__init__()
self.out_dim = out_dim
# 第一层,3个卷积层和一个最大池化层
self.layer1 = nn.Sequential(
# 输入1通道,卷积核长度为3,输出32通道(如192的向量,(192+2*1-3)/2+1=96,输出96*32)
nn.Conv1d(1, 32, 3, stride=2, padding=1),
nn.BatchNorm1d(32),
nn.ReLU(inplace=True),
# 输入32通道,卷积核3,输出32通道(输入96*32,卷积3*32*32,输出96*32)
nn.Conv1d(32, 32, 3, padding=1),
nn.BatchNorm1d(32),
nn.ReLU(inplace=True),
# 输入32通道,卷积核3,输出32通道(输入96*32,卷积3*32*32,输出96*32)
nn.Conv1d(32, 32, 3, padding=1),
nn.BatchNorm1d(32),
nn.ReLU(inplace=True),
# 输入96*32,输出48*32
nn.MaxPool1d(kernel_size=2, stride=2)
)
# 第二层,2个卷积层和一个最大池化层
self.layer2 = nn.Sequential(
# 输入32通道,卷积核长度为3,输出64通道(输入48*32,卷积3*32*64,输出24*64)
nn.Conv1d(32, 64, 3, stride=2, padding=1),
nn.BatchNorm1d(64),
nn.ReLU(inplace=True),
# 输入64通道,卷积核3,输出64通道(输入24*64,卷积3*64*64,输出24*64)
nn.Conv1d(64, 64, 3, padding=1),
nn.BatchNorm1d(64),
nn.ReLU(inplace=True),
# 输入24*32,输出12*32
nn.MaxPool1d(kernel_size=2, stride=2)
)
# 第三层,2个卷积层和一个最大池化层
self.layer3 = nn.Sequential(
# 输入64通道,卷积核3,输出128通道(输入12*64,卷积3*64*128,输出6*128)
nn.Conv1d(64, 128, 3, stride=2, padding=1),
nn.BatchNorm1d(128),
nn.ReLU(inplace=True),
# 输入128通道,卷积核3,输出128通道(输入6*128,卷积3*128*128,输出6*128)
nn.Conv1d(128, 128, 3, padding=1),
nn.BatchNorm1d(128),
nn.ReLU(inplace=True),
# 输入6*128,输出3*128
nn.MaxPool1d(kernel_size=2, stride=2)
)
self.conv_layer = nn.Sequential(
self.layer1,
self.layer2,
self.layer3
)
self.fc = nn.Sequential(
nn.Linear(3 * 128, 1024),
nn.ReLU(inplace=True),
nn.Dropout(),
nn.Linear(1024, self.out_dim)
)
def forward(self, x):
x = self.conv_layer(x)
x = x.view(-1, 3 * 128)
x = self.fc(x)
return x
class SunQuery(nn.Module):
def __init__(self, out_dim):
super(SunQuery, self).__init__()
self.out_dim = out_dim
# 第一层,3个卷积层和一个最大池化层
self.layer1 = nn.Sequential(
# 输入1通道,卷积核长度为3,输出32通道(如512的向量,(512+2*1-3)/2+1=256,输出256*32)
nn.Conv1d(1, 32, 3, stride=2, padding=1),
nn.BatchNorm1d(32),
nn.ReLU(inplace=True),
# 输入32通道,卷积核3,输出32通道(输入256*32,卷积3*32*32,输出256*32)
nn.Conv1d(32, 32, 3, padding=1),
nn.BatchNorm1d(32),
nn.ReLU(inplace=True),
# 输入32通道,卷积核3,输出32通道(输入256*32,卷积3*32*32,输出256*32)
nn.Conv1d(32, 32, 3, padding=1),
nn.BatchNorm1d(32),
nn.ReLU(inplace=True),
# 输入256*32,输出128*32
nn.MaxPool1d(kernel_size=2, stride=2)
)
# 第二层,2个卷积层和一个最大池化层
self.layer2 = nn.Sequential(
# 输入32通道,卷积核长度为3,输出64通道(输入128*32,卷积3*32*64,输出64*64)
nn.Conv1d(32, 64, 3, stride=2, padding=1),
nn.BatchNorm1d(64),
nn.ReLU(inplace=True),
# 输入64通道,卷积核3,输出64通道(输入64*64,卷积3*64*64,输出64*64)
nn.Conv1d(64, 64, 3, padding=1),
nn.BatchNorm1d(64),
nn.ReLU(inplace=True),
# 输入64*64,输出32*64
nn.MaxPool1d(kernel_size=2, stride=2)
)
# 第三层,3个卷积层和一个最大池化层
self.layer3 = nn.Sequential(
# 输入64通道,卷积核3,输出128通道(输入32*64,卷积3*64*128,输出16*128)
nn.Conv1d(64, 128, 3, stride=2, padding=1),
nn.BatchNorm1d(128),
nn.ReLU(inplace=True),
# 输入128通道,卷积核3,输出128通道(输入16*128,卷积3*128*128,输出16*128)
nn.Conv1d(128, 128, 3, padding=1),
nn.BatchNorm1d(128),
nn.ReLU(inplace=True),
# 输入128通道,卷积核3,输出128通道(输入16*128,卷积3*128*128,输出16*128)
nn.Conv1d(128, 128, 3, padding=1),
nn.BatchNorm1d(128),
nn.ReLU(inplace=True),
# 输入16*128,输出8*128
nn.MaxPool1d(kernel_size=2, stride=2)
)
self.conv_layer = nn.Sequential(
self.layer1,
self.layer2,
self.layer3
)
self.fc = nn.Sequential(
nn.Linear(8 * 128, 1024),
nn.ReLU(inplace=True),
nn.Dropout(),
nn.Linear(1024, self.out_dim)
)
def forward(self, x):
x = self.conv_layer(x)
x = x.view(-1, 8 * 128)
x = self.fc(x)
return x
class EnronQuery(nn.Module):
def __init__(self, out_dim):
super(EnronQuery, self).__init__()
self.out_dim = out_dim
# 第一层,3个卷积层和一个最大池化层
self.layer1 = nn.Sequential(
# 输入1通道,卷积核长度为3,输出32通道(如1369的向量,(1369+2*1-3)/2+1=685,输出685*32)
nn.Conv1d(1, 32, 3, stride=2, padding=1),
nn.BatchNorm1d(32),
nn.ReLU(inplace=True),
# 输入32通道,卷积核3,输出32通道(输入685*32,卷积3*32*32,输出685*32)
nn.Conv1d(32, 32, 3, padding=1),
nn.BatchNorm1d(32),
nn.ReLU(inplace=True),
# 输入32通道,卷积核3,输出32通道(输入685*32,卷积3*32*32,输出685*32)
nn.Conv1d(32, 32, 3, padding=1),
nn.BatchNorm1d(32),
nn.ReLU(inplace=True),
# 输入685*32,输出342*32
nn.MaxPool1d(kernel_size=2, stride=2)
)
# 第二层,2个卷积层和一个最大池化层
self.layer2 = nn.Sequential(
# 输入32通道,卷积核长度为3,输出64通道(输入342*32,卷积3*32*64,输出171*64)
nn.Conv1d(32, 64, 3, stride=2, padding=1),
nn.BatchNorm1d(64),
nn.ReLU(inplace=True),
# 输入64通道,卷积核3,输出64通道(输入171*64,卷积3*64*64,输出171*64)
nn.Conv1d(64, 64, 3, padding=1),
nn.BatchNorm1d(64),
nn.ReLU(inplace=True),
# 输入171*64,输出85*64
nn.MaxPool1d(kernel_size=2, stride=2)
)
# 第三层,3个卷积层和一个最大池化层
self.layer3 = nn.Sequential(
# 输入64通道,卷积核3,输出128通道(输入85*64,卷积3*64*128,输出43*128)
nn.Conv1d(64, 128, 3, stride=2, padding=1),
nn.BatchNorm1d(128),
nn.ReLU(inplace=True),
# 输入128通道,卷积核3,输出128通道(输入43*128,卷积3*128*128,输出43*128)
nn.Conv1d(128, 128, 3, padding=1),
nn.BatchNorm1d(128),
nn.ReLU(inplace=True),
# 输入128通道,卷积核3,输出128通道(输入43*128,卷积3*128*128,输出43*128)
nn.Conv1d(128, 128, 3, padding=1),
nn.BatchNorm1d(128),
nn.ReLU(inplace=True),
# 输入43*128,输出21*128
nn.MaxPool1d(kernel_size=2, stride=2)
)
# 第四层,3个卷积层和一个最大池化层
self.layer4 = nn.Sequential(
# 输入128通道,卷积核3,输出256通道(输入21*128,卷积3*128*256,输出11*256)
nn.Conv1d(128, 256, 3, stride=2, padding=1),
nn.BatchNorm1d(256),
nn.ReLU(inplace=True),
# 输入256通道,卷积核3,输出256通道(输入11*256,卷积3*256*256,输出11*256)
nn.Conv1d(256, 256, 3, padding=1),
nn.BatchNorm1d(256),
nn.ReLU(inplace=True),
# 输入256通道,卷积核3,输出256通道(输入11*256,卷积3*256*256,输出11*256)
nn.Conv1d(256, 256, 3, padding=1),
nn.BatchNorm1d(256),
nn.ReLU(inplace=True),
# 输入11*256,输出5*256
nn.MaxPool1d(kernel_size=2, stride=2)
)
self.conv_layer = nn.Sequential(
self.layer1,
self.layer2,
self.layer3,
self.layer4
)
self.fc = nn.Sequential(
nn.Linear(5 * 256, 1024),
nn.ReLU(inplace=True),
nn.Dropout(),
nn.Linear(1024, self.out_dim)
)
def forward(self, x):
x = self.conv_layer(x)
x = x.view(-1, 5 * 256)
x = self.fc(x)
return x
class NuswideQuery(nn.Module):
def __init__(self, out_dim):
super(NuswideQuery, self).__init__()
self.out_dim = out_dim
# 第一层,3个卷积层和一个最大池化层
self.layer1 = nn.Sequential(
# 输入1通道,卷积核长度为3,输出32通道(如500的向量,(500+2*1-3)/2+1=250,输出250*32)
nn.Conv1d(1, 32, 3, stride=2, padding=1),
nn.BatchNorm1d(32),
nn.ReLU(inplace=True),
# 输入32通道,卷积核3,输出32通道(输入250*32,卷积3*32*32,输出250*32)
nn.Conv1d(32, 32, 3, padding=1),
nn.BatchNorm1d(32),
nn.ReLU(inplace=True),
# 输入32通道,卷积核3,输出32通道(输入250*32,卷积3*32*32,输出250*32)
nn.Conv1d(32, 32, 3, padding=1),
nn.BatchNorm1d(32),
nn.ReLU(inplace=True),
# 输入250*32,输出125*32
nn.MaxPool1d(kernel_size=2, stride=2)
)
# 第二层,2个卷积层和一个最大池化层
self.layer2 = nn.Sequential(
# 输入32通道,卷积核长度为3,输出64通道(输入125*32,卷积3*32*64,输出63*64)
nn.Conv1d(32, 64, 3, stride=2, padding=1),
nn.BatchNorm1d(64),
nn.ReLU(inplace=True),
# 输入64通道,卷积核3,输出64通道(输入63*64,卷积3*64*64,输出63*64)
nn.Conv1d(64, 64, 3, padding=1),
nn.BatchNorm1d(64),
nn.ReLU(inplace=True),
# 输入63*64,输出31*64
nn.MaxPool1d(kernel_size=2, stride=2)
)
# 第三层,3个卷积层和一个最大池化层
self.layer3 = nn.Sequential(
# 输入64通道,卷积核3,输出128通道(输入31*64,卷积3*64*128,输出16*128)
nn.Conv1d(64, 128, 3, stride=2, padding=1),
nn.BatchNorm1d(128),
nn.ReLU(inplace=True),
# 输入128通道,卷积核3,输出128通道(输入16*128,卷积3*128*128,输出16*128)
nn.Conv1d(128, 128, 3, padding=1),
nn.BatchNorm1d(128),
nn.ReLU(inplace=True),
# 输入128通道,卷积核3,输出128通道(输入16*128,卷积3*128*128,输出16*128)
nn.Conv1d(128, 128, 3, padding=1),
nn.BatchNorm1d(128),
nn.ReLU(inplace=True),
# 输入16*128,输出8*128
nn.MaxPool1d(kernel_size=2, stride=2)
)
self.conv_layer = nn.Sequential(
self.layer1,
self.layer2,
self.layer3
)
self.fc = nn.Sequential(
nn.Linear(8 * 128, 1024),
nn.ReLU(inplace=True),
nn.Dropout(),
nn.Linear(1024, self.out_dim)
)
def forward(self, x):
x = self.conv_layer(x)
x = x.view(-1, 8 * 128)
x = self.fc(x)
return x
class NotreQuery(nn.Module):
def __init__(self, out_dim):
super(NotreQuery, self).__init__()
self.out_dim = out_dim
# 第一层,2个卷积层和一个最大池化层
self.layer1 = nn.Sequential(
# 输入1通道,卷积核长度为3,输出32通道(如128的向量,(128+2*1-3)/2+1=64,输出64*32)
nn.Conv1d(1, 32, 3, stride=2, padding=1),
nn.BatchNorm1d(32),
nn.ReLU(inplace=True),
# 输入32通道,卷积核3,输出32通道(输入64*32,卷积3*32*32,输出64*32)
nn.Conv1d(32, 32, 3, padding=1),
nn.BatchNorm1d(32),
nn.ReLU(inplace=True),
# 输入32通道,卷积核3,输出32通道(输入64*32,卷积3*32*32,输出64*32)
nn.Conv1d(32, 32, 3, padding=1),
nn.BatchNorm1d(32),
nn.ReLU(inplace=True),
# 输入64*32,输出32*32
nn.MaxPool1d(kernel_size=2, stride=2)
)
# 第二层,2个卷积层和一个最大池化层
self.layer2 = nn.Sequential(
# 输入32通道,卷积核长度为3,输出65通道(输入32*32,卷积3*32*64,输出16*64)
nn.Conv1d(32, 64, 3, stride=2, padding=1),
nn.BatchNorm1d(64),
nn.ReLU(inplace=True),
# 输入64通道,卷积核3,输出64通道(输入16*64,卷积3*64*64,输出16*64)
nn.Conv1d(64, 64, 3, padding=1),
nn.BatchNorm1d(64),
nn.ReLU(inplace=True),
# 输入16*64,输出8*64
nn.MaxPool1d(kernel_size=2, stride=2)
)
# 第三层,3个卷积层和一个最大池化层
self.layer3 = nn.Sequential(
# 输入64通道,卷积核3,输出128通道(输入8*64,卷积3*64*128,输出4*128)
nn.Conv1d(64, 128, 3, stride=2, padding=1),
nn.BatchNorm1d(128),
nn.ReLU(inplace=True),
# 输入128通道,卷积核3,输出128通道(输入4*128,卷积3*128*128,输出4*128)
nn.Conv1d(128, 128, 3, padding=1),
nn.BatchNorm1d(128),
nn.ReLU(inplace=True),
# 输入128通道,卷积核3,输出128通道(输入4*128,卷积3*128*128,输出4*128)
nn.Conv1d(128, 128, 3, padding=1),
nn.BatchNorm1d(128),
nn.ReLU(inplace=True),
# 输入4*128,输出2*128
nn.MaxPool1d(kernel_size=2, stride=2)
)
self.conv_layer = nn.Sequential(
self.layer1,
self.layer2,
self.layer3
)
self.fc = nn.Sequential(
nn.Linear(2 * 128, 1024),
nn.ReLU(inplace=True),
nn.Dropout(),
nn.Linear(1024, self.out_dim)
)
def forward(self, x):
x = self.conv_layer(x)
x = x.view(-1, 2 * 128)
x = self.fc(x)
return x
if __name__ == '__main__':
x = torch.Tensor([i for i in range(128)])
x = x.unsqueeze(0)
x = x.unsqueeze(0)
net = NotreQuery(2000)
x = net(x)
print(x.shape)