-
Notifications
You must be signed in to change notification settings - Fork 0
/
h5_load.py
412 lines (367 loc) · 17.4 KB
/
h5_load.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
#coding: utf-8
#------------------------------------------------------------------------
# Description: load keras model weights HDF5 file
# Date: 2018.6
#------------------------------------------------------------------------
import h5py
import numpy as np
# check version
# python 3.6.4 (64bit) on win32
# windows 10 (64bit)
# numpy (1.14.0)
# h5py (2.7.1)
def irekae4(w1):
# coefficient position changeconv (*,*,a,a)->(*.*,a',a')
w2=np.zeros( w1.shape, dtype=np.float32)
for k in range(w1.shape[0]):
for l in range(w1.shape[1]):
for i in range(w1.shape[2]):
for j in range(w1.shape[3]):
w2[k,l,i,j]=w1[k,l,(w1.shape[2]-1)-i, (w1.shape[3]-1)-j]
return w2
class Class_net_from_h5_CNN(object):
def __init__(self, IN_FILE='data/music_tagger_cnn_weights_theano.h5'):
"""
Download music-auto_tagging-keras's CNN weights file of
<https://github.com/keunwoochoi/music-auto_tagging-keras/tree/master/data/music_tagger_cnn_weights_theano.h5>
and put it in the 'data/' directory.
Or,
Download music-auto_tagging-keras's CNN weights file of
<https://github.com/keunwoochoi/music-auto_tagging-keras/tree/90b294091adaada477b9003201dd20a9fe15a3c1/data/cnn_weights_theano.h5>
and put it in the 'data/' directory.
Contents:
IN_FILE='data/music_tagger_cnn_weights_theano.h5'
['bn1', 'bn2', 'bn3', 'bn4', 'bn5', 'bn_0_freq',
'conv1', 'conv2', 'conv3', 'conv4', 'conv5',
'dropout1', 'dropout2', 'dropout3', 'dropout4', 'dropout5',
'elu_1', 'elu_2', 'elu_3', 'elu_4', 'elu_5',
'flatten_1', 'input_1', 'output',
'pool1', 'pool2', 'pool3', 'pool4', 'pool5']
bn1 ['bn1_beta', 'bn1_gamma', 'bn1_running_mean', 'bn1_running_std']
conv1 ['conv1_W', 'conv1_b']
output ['output_W', 'output_b']
"""
self.IN_FILE = IN_FILE
self.model_weights = h5py.File(self.IN_FILE, 'r')
@property
def norm0_b(self):
return self.model_weights['bn_0_freq/bn_0_freq_beta'].value.reshape(self.model_weights['bn_0_freq/bn_0_freq_beta'].value.size)
@property
def norm0_g(self):
return self.model_weights['bn_0_freq/bn_0_freq_gamma'].value.reshape(self.model_weights['bn_0_freq/bn_0_freq_gamma'].value.size)
@property
def norm0_m(self):
return self.model_weights['bn_0_freq/bn_0_freq_running_mean'].value.reshape(self.model_weights['bn_0_freq/bn_0_freq_running_mean'].value.size)
@property
def norm0_v(self):
return self.model_weights['bn_0_freq/bn_0_freq_running_std'].value.reshape(self.model_weights['bn_0_freq/bn_0_freq_running_std'].value.size)
@property
def conv1_W(self):
return irekae4(self.model_weights['conv1/conv1_W'].value)
@property
def conv1_b(self):
return self.model_weights['conv1/conv1_b'].value.reshape(self.model_weights['conv1/conv1_b'].value.size)
@property
def norm1_b(self):
return self.model_weights['bn1/bn1_beta'].value.reshape(self.model_weights['bn1/bn1_beta'].value.size)
@property
def norm1_g(self):
return self.model_weights['bn1/bn1_gamma'].value.reshape(self.model_weights['bn1/bn1_gamma'].value.size)
@property
def norm1_m(self):
return self.model_weights['bn1/bn1_running_mean'].value.reshape(self.model_weights['bn1/bn1_running_mean'].value.size)
@property
def norm1_v(self):
return self.model_weights['bn1/bn1_running_std'].value.reshape(self.model_weights['bn1/bn1_running_std'].value.size)
@property
def conv2_W(self):
return irekae4(self.model_weights['conv2/conv2_W'].value)
@property
def conv2_b(self):
return self.model_weights['conv2/conv2_b'].value.reshape(self.model_weights['conv2/conv2_b'].value.size)
@property
def norm2_b(self):
return self.model_weights['bn2/bn2_beta'].value.reshape(self.model_weights['bn2/bn2_beta'].value.size)
@property
def norm2_g(self):
return self.model_weights['bn2/bn2_gamma'].value.reshape(self.model_weights['bn2/bn2_gamma'].value.size)
@property
def norm2_m(self):
return self.model_weights['bn2/bn2_running_mean'].value.reshape(self.model_weights['bn2/bn2_running_mean'].value.size)
@property
def norm2_v(self):
return self.model_weights['bn2/bn2_running_std'].value.reshape(self.model_weights['bn2/bn2_running_std'].value.size)
@property
def conv3_W(self):
return irekae4(self.model_weights['conv3/conv3_W'].value)
@property
def conv3_b(self):
return self.model_weights['conv3/conv3_b'].value.reshape(self.model_weights['conv3/conv3_b'].value.size)
@property
def norm3_b(self):
return self.model_weights['bn3/bn3_beta'].value.reshape(self.model_weights['bn3/bn3_beta'].value.size)
@property
def norm3_g(self):
return self.model_weights['bn3/bn3_gamma'].value.reshape(self.model_weights['bn3/bn3_gamma'].value.size)
@property
def norm3_m(self):
return self.model_weights['bn3/bn3_running_mean'].value.reshape(self.model_weights['bn3/bn3_running_mean'].value.size)
@property
def norm3_v(self):
return self.model_weights['bn3/bn3_running_std'].value.reshape(self.model_weights['bn3/bn3_running_std'].value.size)
@property
def conv4_W(self):
return irekae4(self.model_weights['conv4/conv4_W'].value)
@property
def conv4_b(self):
return self.model_weights['conv4/conv4_b'].value.reshape(self.model_weights['conv4/conv4_b'].value.size)
@property
def norm4_b(self):
return self.model_weights['bn4/bn4_beta'].value.reshape(self.model_weights['bn4/bn4_beta'].value.size)
@property
def norm4_g(self):
return self.model_weights['bn4/bn4_gamma'].value.reshape(self.model_weights['bn4/bn4_gamma'].value.size)
@property
def norm4_m(self):
return self.model_weights['bn4/bn4_running_mean'].value.reshape(self.model_weights['bn4/bn4_running_mean'].value.size)
@property
def norm4_v(self):
return self.model_weights['bn4/bn4_running_std'].value.reshape(self.model_weights['bn4/bn4_running_std'].value.size)
@property
def conv5_W(self):
return irekae4(self.model_weights['conv5/conv5_W'].value)
@property
def conv5_b(self):
return self.model_weights['conv5/conv5_b'].value.reshape(self.model_weights['conv5/conv5_b'].value.size)
@property
def norm5_b(self):
return self.model_weights['bn5/bn5_beta'].value.reshape(self.model_weights['bn5/bn5_beta'].value.size)
@property
def norm5_g(self):
return self.model_weights['bn5/bn5_gamma'].value.reshape(self.model_weights['bn5/bn5_gamma'].value.size)
@property
def norm5_m(self):
return self.model_weights['bn5/bn5_running_mean'].value.reshape(self.model_weights['bn5/bn5_running_mean'].value.size)
@property
def norm5_v(self):
return self.model_weights['bn5/bn5_running_std'].value.reshape(self.model_weights['bn5/bn5_running_std'].value.size)
@property
def fc1_W(self):
# output tag 50
self.tags=50
return self.model_weights['output/output_W'].value.T
@property
def fc1_b(self):
# output tag 50
return self.model_weights['output/output_b'].value.reshape(self.model_weights['output/output_b'].value.size)
class Class_net_from_h5_CRNN(object):
def __init__(self, IN_FILE='data/crnn_net_gru_adam_ours_epoch_40.h5'):
""""
Download Music-Genre-Classification-with-Deep-Learning's CRNN weights file of
<https://github.com/jsalbert/Music-Genre-Classification-with-Deep-Learning/tree/master/models_trained/example_model/weights/crnn_net_gru_adam_ours_epoch_40.h5>
And put it in the 'data/' directory.
Or,
Download music-auto_tagging-keras's CRNN weights file of
<https://github.com/keunwoochoi/music-auto_tagging-keras/tree/master/data/music_tagger_crnn_weights_theano.h5>
and put it in the 'data/' directory.
Contents:
IN_FILE='data/crnn_net_gru_adam_ours_epoch_40.h5'
['bn1', 'bn2', 'bn3', 'bn4', 'bn_0_freq',
'conv1', 'conv2', 'conv3', 'conv4',
'dropout1', 'dropout2', 'dropout3', 'dropout4',
'elu_1', 'elu_2', 'elu_3', 'elu_4', 'final_drop', 'gru1', 'gru2',
'input_1', 'permute_1', 'pool1', 'pool2', 'pool3', 'pool4', 'preds',
'reshape_1', 'zeropadding2d_1']
IN_FILE='data/music_tagger_crnn_weights_theano.h5'
['bn1', 'bn2', 'bn3', 'bn4', 'bn_0_freq',
'conv1', 'conv2', 'conv3', 'conv4',
'dropout1', 'dropout2', 'dropout3', 'dropout4', 'dropout_1',
'elu_6', 'elu_7', 'elu_8', 'elu_9',
'gru1', 'gru2', 'input_2', 'output', 'permute_1',
'pool1', 'pool2', 'pool3', 'pool4', 'reshape_1', 'zeropadding2d_1']
['bn1_beta', 'bn1_gamma', 'bn1_running_mean', 'bn1_running_std']
['bn_0_freq_beta', 'bn_0_freq_gamma', 'bn_0_freq_running_mean', 'bn_0_freq_running_std']
['conv1_W', 'conv1_b']
gru1 ['gru1_U_h', 'gru1_U_r', 'gru1_U_z',
'gru1_W_h', 'gru1_W_r', 'gru1_W_z',
'gru1_b_h', 'gru1_b_r', 'gru1_b_z']
output ['output_W', 'output_b']
"""
self.IN_FILE = IN_FILE
self.model_weights = h5py.File(self.IN_FILE, 'r')
@property
def norm0_b(self):
return self.model_weights['bn_0_freq/bn_0_freq_beta'].value.reshape(self.model_weights['bn_0_freq/bn_0_freq_beta'].value.size)
@property
def norm0_g(self):
return self.model_weights['bn_0_freq/bn_0_freq_gamma'].value.reshape(self.model_weights['bn_0_freq/bn_0_freq_gamma'].value.size)
@property
def norm0_m(self):
return self.model_weights['bn_0_freq/bn_0_freq_running_mean'].value.reshape(self.model_weights['bn_0_freq/bn_0_freq_running_mean'].value.size)
@property
def norm0_v(self):
return self.model_weights['bn_0_freq/bn_0_freq_running_std'].value.reshape(self.model_weights['bn_0_freq/bn_0_freq_running_std'].value.size)
@property
def conv1_W(self):
return irekae4( self.model_weights['conv1/conv1_W'].value)
@property
def conv1_b(self):
return self.model_weights['conv1/conv1_b'].value.reshape(self.model_weights['conv1/conv1_b'].value.size)
@property
def norm1_b(self):
return self.model_weights['bn1/bn1_beta'].value.reshape(self.model_weights['bn1/bn1_beta'].value.size)
@property
def norm1_g(self):
return self.model_weights['bn1/bn1_gamma'].value.reshape(self.model_weights['bn1/bn1_gamma'].value.size)
@property
def norm1_m(self):
return self.model_weights['bn1/bn1_running_mean'].value.reshape(self.model_weights['bn1/bn1_running_mean'].value.size)
@property
def norm1_v(self):
return self.model_weights['bn1/bn1_running_std'].value.reshape(self.model_weights['bn1/bn1_running_std'].value.size)
@property
def conv2_W(self):
return irekae4(self.model_weights['conv2/conv2_W'].value)
@property
def conv2_b(self):
return self.model_weights['conv2/conv2_b'].value.reshape(self.model_weights['conv2/conv2_b'].value.size)
@property
def norm2_b(self):
return self.model_weights['bn2/bn2_beta'].value.reshape(self.model_weights['bn2/bn2_beta'].value.size)
@property
def norm2_g(self):
return self.model_weights['bn2/bn2_gamma'].value.reshape(self.model_weights['bn2/bn2_gamma'].value.size)
@property
def norm2_m(self):
return self.model_weights['bn2/bn2_running_mean'].value.reshape(self.model_weights['bn2/bn2_running_mean'].value.size)
@property
def norm2_v(self):
return self.model_weights['bn2/bn2_running_std'].value.reshape(self.model_weights['bn2/bn2_running_std'].value.size)
@property
def conv3_W(self):
return irekae4(self.model_weights['conv3/conv3_W'].value)
@property
def conv3_b(self):
return self.model_weights['conv3/conv3_b'].value.reshape(self.model_weights['conv3/conv3_b'].value.size)
@property
def norm3_b(self):
return self.model_weights['bn3/bn3_beta'].value.reshape(self.model_weights['bn3/bn3_beta'].value.size)
@property
def norm3_g(self):
return self.model_weights['bn3/bn3_gamma'].value.reshape(self.model_weights['bn3/bn3_gamma'].value.size)
@property
def norm3_m(self):
return self.model_weights['bn3/bn3_running_mean'].value.reshape(self.model_weights['bn3/bn3_running_mean'].value.size)
@property
def norm3_v(self):
return self.model_weights['bn3/bn3_running_std'].value.reshape(self.model_weights['bn3/bn3_running_std'].value.size)
@property
def conv4_W(self):
return irekae4(self.model_weights['conv4/conv4_W'].value)
@property
def conv4_b(self):
return self.model_weights['conv4/conv4_b'].value.reshape(self.model_weights['conv4/conv4_b'].value.size)
@property
def norm4_b(self):
return self.model_weights['bn4/bn4_beta'].value.reshape(self.model_weights['bn4/bn4_beta'].value.size)
@property
def norm4_g(self):
return self.model_weights['bn4/bn4_gamma'].value.reshape(self.model_weights['bn4/bn4_gamma'].value.size)
@property
def norm4_m(self):
return self.model_weights['bn4/bn4_running_mean'].value.reshape(self.model_weights['bn4/bn4_running_mean'].value.size)
@property
def norm4_v(self):
return self.model_weights['bn4/bn4_running_std'].value.reshape(self.model_weights['bn4/bn4_running_std'].value.size)
@property
def gru1_W_r(self):
return self.model_weights['gru1/gru1_W_r'].value.T
@property
def gru1_W_z(self):
return self.model_weights['gru1/gru1_W_z'].value.T
@property
def gru1_W(self):
return self.model_weights['gru1/gru1_W_h'].value.T
@property
def gru1_b_r(self):
return self.model_weights['gru1/gru1_b_r'].value.reshape(self.model_weights['gru1/gru1_b_r'].value.size)
@property
def gru1_b_z(self):
return self.model_weights['gru1/gru1_b_z'].value.reshape(self.model_weights['gru1/gru1_b_z'].value.size)
@property
def gru1_b(self):
return self.model_weights['gru1/gru1_b_h'].value.reshape(self.model_weights['gru1/gru1_b_h'].value.size)
@property
def gru1_U_r(self):
return self.model_weights['gru1/gru1_U_r'].value.T
@property
def gru1_U_z(self):
return self.model_weights['gru1/gru1_U_z'].value.T
@property
def gru1_U(self):
return self.model_weights['gru1/gru1_U_h'].value.T
@property
def gru2_W_r(self):
return self.model_weights['gru2/gru2_W_r'].value.T
@property
def gru2_W_z(self):
return self.model_weights['gru2/gru2_W_z'].value.T
@property
def gru2_W(self):
return self.model_weights['gru2/gru2_W_h'].value.T
@property
def gru2_b_r(self):
return self.model_weights['gru2/gru2_b_r'].value.reshape(self.model_weights['gru2/gru2_b_r'].value.size)
@property
def gru2_b_z(self):
return self.model_weights['gru2/gru2_b_z'].value.reshape(self.model_weights['gru2/gru2_b_z'].value.size)
@property
def gru2_b(self):
return self.model_weights['gru2/gru2_b_h'].value.reshape(self.model_weights['gru2/gru2_b_h'].value.size)
@property
def gru2_U_r(self):
return self.model_weights['gru2/gru2_U_r'].value.T
@property
def gru2_U_z(self):
return self.model_weights['gru2/gru2_U_z'].value.T
@property
def gru2_U(self):
return self.model_weights['gru2/gru2_U_h'].value.T
@property
def fc1_W(self):
if self.IN_FILE == "data/music_tagger_crnn_weights_theano.h5":
# output tag 50
self.tags=50
return self.model_weights['output/output_W'].value.T
else:
# output tag 10
self.tags=10
return self.model_weights['preds/preds_W'].value.T
@property
def fc1_b(self):
if self.IN_FILE == "data/music_tagger_crnn_weights_theano.h5":
# output tag 50
print ('change model to output is 50')
return self.model_weights['output/output_b'].value.reshape(self.model_weights['output/output_b'].value.size)
else:
# output tag 10
return self.model_weights['preds/preds_b'].value.reshape(self.model_weights['preds/preds_b'].value.size)
if __name__ == '__main__':
IN_FILE='data/crnn_net_gru_adam_ours_epoch_40.h5'
model_weights = h5py.File(IN_FILE, 'r')
#print (model_weights.keys())
print (list(model_weights)) # abc order
print (list(model_weights['bn_0_freq'].keys()))
print (list(model_weights['conv1'].keys()))
print (list(model_weights['gru1'].keys()))
print (list(model_weights['bn1'].keys()))
print ('CRNN; bn_1: mean, var, gamma, beta')
a=model_weights['bn1/bn1_running_mean'].value
b=model_weights['bn1/bn1_running_std'].value
c=model_weights['bn1/bn1_gamma'].value
d=model_weights['bn1/bn1_beta'].value
print ('shape',a.shape)
print (a)
print (b)
print (c)
print (d)