/
data_generator.py
241 lines (166 loc) · 6.49 KB
/
data_generator.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
from __future__ import print_function
import os
import numpy as np
import random
import threading
import argparse
class threadsafe_iter:
def __init__(self,it):
self.it = it
self.lock = threading.Lock()
def __iter__(self):
return self
def next(self):
with self.lock:
return self.it.next()
def threadsafe_generator(f):
def g(*a, **kw):
return threadsafe_iter(f(*a,**kw))
return g
@threadsafe_generator
def train_generator_siamese(train_list,y_train,artist_train,mel_mean,mel_std,N_negs,steps_per_epoch,batch_size,feature_path,num_frame_input):
# shuffling pos_anchor pos_item neg_items, audio 129 frames, random selection
while True:
for batch_iter in range(0, steps_per_epoch*batch_size, batch_size):
# initialization
x_train_batch = []
pos_anchor_x_train_batch = []
pos_item_x_train_batch = []
neg_items_x_train_batch = []
for item_idx,item_iter in enumerate(range(batch_iter,batch_iter+batch_size)):
# pos anchor
file_path = feature_path + train_list[item_iter] + '.npy'
tmp = np.load(file_path)
tmp = tmp.T
tmp -= mel_mean
tmp /= mel_std
start = random.randint(0,tmp.shape[0]-num_frame_input)
pos_anchor_x_train = tmp[start:start+num_frame_input,:]
# pos item
#print(artist_train,item_iter)
pos_items = artist_train[y_train[item_iter]]
pos_items_candidate = list(set(pos_items) - set([train_list[item_iter]]))
pos_item = random.choice(pos_items_candidate)
file_path = feature_path + pos_item + '.npy'
tmp = np.load(file_path)
tmp = tmp.T
tmp -= mel_mean
tmp /= mel_std
start = random.randint(0,tmp.shape[0]-num_frame_input)
pos_item_x_train = tmp[start:start+num_frame_input,:]
# neg items
neg_items_candidate = list(set(train_list) - set(pos_items))
random.shuffle(neg_items_candidate)
neg_items = neg_items_candidate[0:N_negs]
neg_items_x_train = [[] for j in range(N_negs)]
for neg_iter in range(N_negs):
file_path = feature_path + neg_items[neg_iter] + '.npy'
tmp = np.load(file_path)
tmp = tmp.T
tmp -= mel_mean
tmp /= mel_std
start = random.randint(0,tmp.shape[0]-num_frame_input)
neg_items_x_train[neg_iter] = tmp[start:start+num_frame_input,:]
pos_anchor_x_train_batch.append(pos_anchor_x_train)
pos_item_x_train_batch.append(pos_item_x_train)
neg_items_x_train_batch.append(neg_items_x_train)
pos_anchor_x_train_batch = np.array(pos_anchor_x_train_batch)
pos_item_x_train_batch = np.array(pos_item_x_train_batch)
neg_items_x_train_batch = np.array(neg_items_x_train_batch)
x_train_batch = [pos_anchor_x_train_batch, pos_item_x_train_batch] + [neg_items_x_train_batch[:,j,:,:] for j in range(N_negs)]
# y_train
y_train_batch = np.zeros((batch_size,N_negs+1))
y_train_batch[:,0] = 1
yield x_train_batch, y_train_batch
def load_valid_siamese(valid_list,y_valid_init,artist_valid,mel_mean,mel_std,N_negs,feature_path,num_frame_input):
# load valid sets
pos_anchor_x_valid = []
pos_item_x_valid = []
neg_items_x_valid = []
for item_iter in range(len(valid_list)):
# pos anchor
file_path = feature_path + valid_list[item_iter] + '.npy'
tmp = np.load(file_path)
tmp = tmp.T
tmp -= mel_mean
tmp /= mel_std
start = random.randint(0,tmp.shape[0]-num_frame_input)
pos_anchor_x_tmp = tmp[start:start+num_frame_input,:]
# pos item
pos_items = artist_valid[y_valid_init[item_iter]]
pos_items_candidate = list(set(pos_items) - set([valid_list[item_iter]]))
pos_item = random.choice(pos_items_candidate)
file_path = feature_path + pos_item + '.npy'
tmp = np.load(file_path)
tmp = tmp.T
tmp -= mel_mean
tmp /= mel_std
start = random.randint(0,tmp.shape[0]-num_frame_input)
pos_item_x_tmp = tmp[start:start+num_frame_input,:]
# neg items
neg_items_candidate = list(set(valid_list) - set(pos_items))
random.shuffle(neg_items_candidate)
neg_items = neg_items_candidate[0:N_negs]
neg_items_x_tmp = [[] for j in range(N_negs)]
for neg_iter in range(N_negs):
file_path = feature_path + neg_items[neg_iter] + '.npy'
tmp = np.load(file_path)
tmp = tmp.T
tmp -= mel_mean
tmp /= mel_std
start = random.randint(0,tmp.shape[0]-num_frame_input)
neg_items_x_tmp[neg_iter] = tmp[start:start+num_frame_input,:]
pos_anchor_x_valid.append(pos_anchor_x_tmp)
pos_item_x_valid.append(pos_item_x_tmp)
neg_items_x_valid.append(neg_items_x_tmp)
if np.remainder(item_iter,1000) == 0:
print(item_iter)
print(item_iter+1)
pos_anchor_x_valid = np.array(pos_anchor_x_valid)
pos_item_x_valid = np.array(pos_item_x_valid)
neg_items_x_valid = np.array(neg_items_x_valid)
x_valid = [pos_anchor_x_valid, pos_item_x_valid] + [neg_items_x_valid[:,j,:,:] for j in range(N_negs)]
# y_train
y_valid = np.zeros((len(valid_list),N_negs+1))
y_valid[:,0] = 1
return x_valid, y_valid
@threadsafe_generator
def train_generator_basic(train_list,y_train_init,artist_train,mel_mean,mel_std,num_sing,steps_per_epoch,batch_size,feature_path,num_frame_input):
# shuffling pos_anchor pos_item neg_items, audio 129 frames, random selection
while True:
for batch_iter in range(0, steps_per_epoch*batch_size, batch_size):
# initialization
x_train_batch = []
y_train_batch = np.zeros((batch_size,num_sing))
for item_idx,item_iter in enumerate(range(batch_iter,batch_iter+batch_size)):
# pos anchor
file_path = feature_path + train_list[item_iter] + '.npy'
tmp = np.load(file_path)
tmp = tmp.T
tmp -= mel_mean
tmp /= mel_std
start = random.randint(0,tmp.shape[0]-num_frame_input)
pos_anchor_x_train = tmp[start:start+num_frame_input,:]
x_train_batch.append(pos_anchor_x_train)
y_train_batch[item_idx,int(y_train_init[item_iter])-1] = 1
x_train_batch = np.array(x_train_batch)
yield x_train_batch, y_train_batch
def load_valid_basic(valid_list,y_valid_init,artist_valid,mel_mean,mel_std,num_sing,feature_path,num_frame_input):
x_valid = []
y_valid = np.zeros((len(valid_list),num_sing))
for item_iter in range(len(valid_list)):
# pos anchor
file_path = feature_path + valid_list[item_iter] + '.npy'
tmp = np.load(file_path)
tmp = tmp.T
tmp -= mel_mean
tmp /= mel_std
start = random.randint(0,tmp.shape[0]-num_frame_input)
pos_anchor_x_tmp = tmp[start:start+num_frame_input,:]
x_valid.append(pos_anchor_x_tmp)
y_valid[item_iter,int(y_valid_init[item_iter])-1] = 1
if np.remainder(item_iter,1000) == 0:
print(item_iter)
print(item_iter)
x_valid = np.array(x_valid)
return x_valid, y_valid