-
Notifications
You must be signed in to change notification settings - Fork 0
/
modules.py
389 lines (349 loc) · 18.3 KB
/
modules.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
from __future__ import print_function
import tensorflow as tf
import numpy as np
import glob
from hyperparams import hyperparams
from networks import encoder, decoder, conv1d, bn, prenet, gru
hp = hyperparams()
"""
1.generator()、discriminator()、
2.get_next_batch()、speaker_embedding()、
1.fast_lstm_3_layers():现在暂时没用了；
"""
def get_next_batch():
"""
这块就是从tfrecord文件中读取已经保存的数据；
"""
# 获取指定目录下的所有tfrecord文件
# #加上r让字符串不转义
tfrecords = glob.glob(f'{hp.TRAIN_DATASET_PATH}/*.tfrecord')
# print("line23: tfrecords = "+str(tfrecords))
"""
tf.train.string_input_producer(
string_tensor,
num_epochs=None, # NUM_EPOCHS = 150；从string_tensor中产生 num_epochs 次字符串；如果未指定，则可以无限次循环遍历字符串
shuffle=True, # shuffle：布尔值。如果为true，则在每个epoch内随机打乱顺序
seed=None,
capacity=32,
shared_name=None,
name=None,
cancel_op=None )
输出字符串到一个输入管道队列
：从TFRecords文件中读取数据， 首先需要用tf.train.string_input_producer（）生成一个解析队列。
之后调用 tf.TFRecordReader 的 tf.parse_single_example 解析器
https://blog.csdn.net/tefuirnever/article/details/90271862
"""
# 输出字符串到一个输入管道队列
filename_queue = tf.train.string_input_producer(tfrecords, shuffle=True, num_epochs=hp.NUM_EPOCHS)
reader = tf.TFRecordReader()
_, serialized_example = reader.read(filename_queue) # 解析器首先读取解析队列，返回serialized_example对象
# 之后调用tf.parse_single_example操作将 Example 协议缓冲区(protocol buffer)解析为张量。
features = tf.parse_single_example(
serialized_example,
features={
'ori_spkid': tf.FixedLenFeature(shape=(1,), dtype=tf.int64),
'ori_mel': tf.VarLenFeature(dtype=tf.float32),
'ori_mel_shape': tf.FixedLenFeature(shape=(2,), dtype=tf.int64),
'aim_spkid': tf.FixedLenFeature(shape=(1,), dtype=tf.int64),
'aim_mel': tf.VarLenFeature(dtype=tf.float32),
'aim_mel_shape': tf.FixedLenFeature(shape=(2,), dtype=tf.int64),
}
)
# tf.sparse_tensor_to_dense 将 SparseTensor 转换为稠密张量.(即理解为，稀疏矩阵，填充上默认值)
features['ori_mel'] = tf.sparse_tensor_to_dense(features['ori_mel'])
features['aim_mel'] = tf.sparse_tensor_to_dense(features['aim_mel'])
ori_spk = features['ori_spkid']
ori_mel = tf.reshape(features['ori_mel'], features['ori_mel_shape'])
aim_spk = features['aim_spkid']
aim_mel = tf.reshape(features['aim_mel'], features['aim_mel_shape'])
# self.CODED_DIM = 60 # 压缩成60维
ori_mel = tf.reshape(ori_mel, [-1, hp.CODED_DIM])
aim_mel = tf.reshape(aim_mel, [-1, hp.CODED_DIM]) # 80 维度 mel
ori_spk_batch, ori_mel_batch, aim_spk_batch, aim_mel_batch = tf.train.batch([ori_spk, ori_mel, aim_spk, aim_mel],
batch_size=hp.BATCH_SIZE,
capacity=100,
num_threads=10,
dynamic_pad=True,
allow_smaller_final_batch=False)
"""
是说在这里，get_next_batch（）函数，返回之前，就可以做 pad 操作吗？
"""
# tf.shape(ori_mel_batch)[1]
max_frame = tf.maximum(tf.shape(ori_mel_batch)[1], tf.shape(aim_mel_batch)[1]) # 最大帧值
gap_frame = max_frame - tf.minimum(tf.shape(ori_mel_batch)[1], tf.shape(aim_mel_batch)[1]) # 帧值 之差
# print(tf.math.subtract(max_frame, tf.shape(aim_mel_batch)[1]))
padded = tf.zeros([tf.shape(aim_mel_batch)[0], tf.subtract(max_frame, tf.shape(aim_mel_batch)[1]),
tf.shape(aim_mel_batch)[2]], dtype=tf.float32)
# a = padded
aim_mel_batch = tf.concat((aim_mel_batch, padded), axis=1)
# concated_1 = aim_mel_batch
padded = tf.zeros([tf.shape(ori_mel_batch)[0], tf.subtract(max_frame, tf.shape(ori_mel_batch)[1]),
tf.shape(ori_mel_batch)[2]], dtype=tf.float32)
# b = padded
# padded = tf.zeros_like([1, tf.math.subtract(max_frame, tf.shape(ori_mel_batch)[1]), 1], dtype=tf.float32)
ori_mel_batch = tf.concat((ori_mel_batch, padded), axis=1)
# concated_2 = ori_mel_batch
# padded = tf.zeros_like([0，差值，0])
# aim_mel_batch = tf.concat((aim_mel_batch, padded), axis=1)
# aim_mel_batch = tf.pad(aim_mel_batch, [[0, 0], [0, tf.math.subtract(max_frame, tf.shape(aim_mel_batch)[1])], [0, 0]], "CONSTANT")
# ori_mel_batch = tf.pad(ori_mel_batch, [[0, 0], [0, tf.math.subtract(max_frame, tf.shape(ori_mel_batch)[1])], [0, 0]], "CONSTANT")
# return ori_spk_batch, ori_mel_batch, aim_spk_batch, aim_mel_batch, a, b,concated_1,concated_2,max_frame
return ori_spk_batch, ori_mel_batch, aim_spk_batch, aim_mel_batch
def speaker_embedding(inputs, spk_num, num_units, zero_pad=True, scope="speaker_embedding", reuse=None):
'''Embeds a given tensor.
输入的inputs，是表示说话人序号，所以是用 int 类型；
num_units = 256，就是把说话人，表示成一个256维度的 embedding 向量
Args:
inputs: A `Tensor` with type `int32` or `int64` containing the ids
to be looked up in `lookup table`.
spk_num: An int. Vocabulary size.
num_units: An int. Number of embedding hidden units.
zero_pad: A boolean. If True, all the values of the fist row (id 0)
should be constant zeros.
scope: Optional scope for `variable_scope`.
reuse: Boolean, whether to reuse the weights of a previous layer
by the same name.
Returns:
A `Tensor` with one more rank than inputs's. The last dimesionality
should be `num_units`.
'''
with tf.variable_scope(scope, reuse=reuse):
lookup_table = tf.get_variable('lookup_table',
dtype=tf.float32,
shape=[spk_num, num_units],
initializer=tf.truncated_normal_initializer(mean=0.0, stddev=0.01))
if zero_pad:
lookup_table = tf.concat((tf.zeros(shape=[1, num_units]),
lookup_table[1:, :]), 0) # [起始行：结束行，起始列：结束列]
return tf.nn.embedding_lookup(lookup_table, inputs)
# 先不用 G 和 D，只用Encoder和Decoder
# 还是先用G，并去掉embeding部分内容就可以了
# def generator(speaker_embedding, inputs, is_training=True, scope_name='generator', reuse=None):
# '''Generate features.
# Args:
# speaker_embedding: A `Tensor` with type `float32` contains speaker information. [N, E]
# inputs: A `Tensor` with type `float32` contains speech features.
# is_training: Boolean, whether to train or inference.
# scope_name: Optional scope for `variable_scope`.
# reuse: Boolean, whether to reuse the weights of a previous layer
# by the same name.
# Returns:
# A decoded `Tensor` with aim speaker.
# vae mu vector.
# vae log_var vector.
# '''
# with tf.variable_scope(scope_name, reuse=reuse):
# sample, mu, log_var = encoder(inputs, is_training=is_training, scope='vae_encoder') # [N, T, E]
# #speaker_embedding = tf.expand_dims(speaker_embedding, axis=1) # [N, 1, E]
# speaker_embedding = tf.tile(speaker_embedding, [1, tf.shape(sample)[1], 1]) # [N, T, E]
# # tf.tile() 用来对张量(Tensor)进行扩展的,表示每一维度，拓展复制几次；
# encoded = tf.concat((speaker_embedding, sample), axis=-1) # [N, T, E+G]
# outputs = decoder(encoded, is_training=is_training, scope='vae_decoder')
# return outputs, mu, log_var # [N, T, C]
# 输入进来 ori_mel 谱【16，N，80】其中 N 和同一batch 的 aim_mel 是一样的
def generator(inputs, is_training=True, scope_name='generator', reuse=None):
with tf.variable_scope(scope_name, reuse=reuse):
sample, mu, log_var = encoder(inputs, is_training=is_training, scope='vae_encoder') # [N, T, E]
# speaker_embedding = tf.tile(speaker_embedding, [1, tf.shape(sample)[1], 1]) # [N, T, E]
# tf.tile() 用来对张量(Tensor)进行扩展的,表示每一维度，拓展复制几次；
# encoded = tf.concat((speaker_embedding, sample), axis=-1) # [N, T, E+G]
outputs = decoder(sample, is_training=is_training, scope='vae_decoder')
return outputs, mu, log_var # [N, T, C]
# 输入：tmp = tf.placeholder(name='ori_feat', shape=[16, 100, 80], dtype=tf.float32)
# 输出：shape=(16, 100, 60)
# 因为CODED_DIM = 60 # 压缩成60维
# (<tf.Tensor 'generator/vae_decoder/dense_2/Tanh:0' shape=(16, 100, 60) dtype=float32>,
# <tf.Tensor 'generator/vae_encoder/mean/BiasAdd:0' shape=(16, 100, 16) dtype=float32>,
# <tf.Tensor 'generator/vae_encoder/log_var/BiasAdd:0' shape=(16, 100, 16) dtype=float32>)
#
# def discriminator(inputs, scope_name='discriminator', reuse=None):
# '''Discriminator features.
#
# Args:
# inputs: A `Tensor` with type `float32` contains speech features. [N, T, F]
# scope_name: Optional scope for `variable_scope`.
# reuse: Boolean, whether to reuse the weights of a previous layer
# by the same name.
#
# Returns:
# A softmax
# '''
# with tf.variable_scope(scope_name, reuse=reuse):
# out = lstm_3_layers(inputs, num_units=hp.CODED_DIM * 2, bidirection=False) # [N, C]
# out = tf.layers.dense(out, units=hp.SPK_NUM * 2, activation=tf.nn.tanh, name='dense1') # [N, L*2]
# out = tf.layers.dense(out, units=hp.SPK_NUM * 2, activation=tf.nn.sigmoid, name='dense2') # [N, L]
# return out
#
#
# def fast_lstm_3_layers(inputs, num_units=None, bidirection=False, scope="lstm_3_layers", reuse=tf.AUTO_REUSE):
# '''
# :param inputs: A 3-d tensor. [N, T, C]
# :param num_units: An integer. The last hidden units.
# :param bidirection: A boolean. If True, bidirectional results are concatenated.
# :param scope: A string. scope name.
# :param reuse: Boolean. whether to reuse the weights of a previous layer.
# :return: if bidirection is True, A 2-d tensor. [N, num_units * 2]
# else, A 2-d tensor. [N, num_units]
# '''
# with tf.variable_scope(scope, reuse=reuse):
# if not num_units:
# num_units = inputs.get_shape().as_list[-1]
# with tf.variable_scope('lstm_1'):
# lstm_1 = tf.keras.layers.CuDNNLSTM(units=num_units, return_sequences=True, return_state=True)
# with tf.variable_scope('lstm_2'):
# lstm_2 = tf.keras.layers.CuDNNLSTM(units=num_units, return_sequences=True, return_state=True)
# with tf.variable_scope('lstm_3'):
# lstm_3 = tf.keras.layers.CuDNNLSTM(units=num_units, return_sequences=False, return_state=True)
# out = lstm_1(inputs)
# out = lstm_2(out[0])
# out = lstm_3(out[0])
# return out[0]
#
#
# def lstm_3_layers(inputs, num_units=None, bidirection=False, scope="lstm", reuse=tf.AUTO_REUSE):
# '''
# :param inputs: A 3-d tensor. [N, T, C]
# :param num_units: An integer. The last hidden units.
# :param bidirection: A boolean. If True, bidirectional results are concatenated.
# :param scope: A string. scope name.
# :param reuse: Boolean. whether to reuse the weights of a previous layer.
# :return: if bidirection is True, A 2-d tensor. [N, num_units * 2]
# else, A 2-d tensor. [N, num_units]
# '''
# with tf.variable_scope(scope, reuse=reuse):
# if not num_units:
# num_units = inputs.get_shape().as_list[-1]
# # cellls = [tf.nn.rnn_cell.LSTMCell(size) for size in [num_units, num_units, num_units]]
# cellls = [tf.nn.rnn_cell.LSTMCell(size) for size in [num_units, num_units, num_units]]
# multi_cell = tf.nn.rnn_cell.MultiRNNCell(cellls)
# if bidirection:
# bw_cells = [tf.nn.rnn_cell.LSTMCell(size) for size in [num_units, num_units, num_units]]
# multi_bw_cell = tf.nn.rnn_cell.MultiRNNCell(bw_cells)
# outputs, final_state = tf.nn.dynamic_rnn(multi_cell, multi_bw_cell, inputs=inputs, dtype=tf.float32)
# # outputs shape : top lstm outputs, ([N, T, num_units], [N, T, num_units])
# # lstm final_state : multi final state stack together, ([N, 2, num_units], [N, 2, num_units])
# return tf.concat(final_state, axis=2)[-1][0]
# outputs, final_state = tf.nn.dynamic_rnn(cell=multi_cell, inputs=inputs, dtype=tf.float32)
# # outputs shape : top lstm outputs, [N, T, num_units]
# # lstm final_state : multi final state stack together, [N, 2, num_units]
# return final_state[-1][0]
# if __name__ == "__main__":
# ori_spk_batch, ori_mel_batch, aim_spk_batch, aim_mel_batch= get_next_batch()
# sess = tf.Session()
# sess.run(tf.initialize_local_variables())
# coord = tf.train.Coordinator()
# threads = tf.train.start_queue_runners(sess, coord)
# try:
# while not coord.should_stop():
# ori_mel, aim_mel = sess.run([ori_mel_batch, aim_mel_batch])
#
# print("1 ori_mel_batch.shape = "+str(ori_mel.shape)+" aim_mel_batch.shape = " + str(aim_mel.shape))
#
#
# except tf.errors.OutOfRangeError:
# print("complete")
# finally:
# coord.request_stop()
# coord.join(threads)
# sess.close()
# tmp = tf.placeholder(name='ori_feat', shape=[16, 100, 80], dtype=tf.float32)
# # print("242")
# # print(generator(tmp))
# (<tf.Tensor 'generator/vae_decoder/dense_2/Tanh:0' shape=(16, 100, 80) dtype=float32>,
# decoder 出来的结果，拿去和aim_mel做对比：【 KL损失 】
# <tf.Tensor 'generator/vae_encoder/mean/BiasAdd:0' shape=(16, 100, 16) dtype=float32>,
# <tf.Tensor 'generator/vae_encoder/log_var/BiasAdd:0' shape=(16, 100, 16) dtype=float32>)
# with tf.Session() as sess:
# sess.run(tf.global_variables_initializer())
# # print(get_next_batch())
# # sess.run(get_next_batch())
# # a, b, c, d = get_next_batch()
# # # tf.print(a)
# # print(sess.run(get_next_batch()))
# a, b, c = sess.run(generator(), feed_dict={inputs: tmp})
# print(a.shape)
# print("end")
# (<tf.Tensor 'batch:0' shape=(16, 1) dtype=int64>,
# <tf.Tensor 'batch:1' shape=(16, ?, 60) dtype=float32>,
# <tf.Tensor 'batch:2' shape=(16, 1) dtype=int64>,
# <tf.Tensor 'batch:3' shape=(16, ?, 60) dtype=float32>)
#
# if __name__ == "__main__":
# ori_spk_batch, ori_mel_batch, aim_spk_batch, aim_mel_batch = get_next_batch()
# sess = tf.Session()
# sess.run(tf.initialize_local_variables())
# coord = tf.train.Coordinator()
# threads = tf.train.start_queue_runners(sess, coord)
# try:
# while not coord.should_stop():
# ori_mel = sess.run(ori_mel_batch)
# aim_mel = sess.run(aim_mel_batch)
# print("1 ori_mel_batch.shape = "+str(ori_mel.shape[1])+" aim_mel_batch.shape = " + str(aim_mel.shape[1]))
#
# max_frame = max(ori_mel.shape[1], aim_mel.shape[1]) # 最大帧值
# gap_frame = max_frame - min(ori_mel.shape[1], aim_mel.shape[1]) # 帧值 之差
# print("max_frame = "+str(max_frame)+" gap_frame = "+str(gap_frame))
#
# # aim_mel = np.pad(aim_mel, ([[0, 0], [0, gap_frame], [0, 0]]), 'constant')
# # def pad(array, pad_width, mode='constant', **kwargs):
# if ori_mel.shape[1] == max_frame:
# aim_mel = np.pad(aim_mel, ([[0, 0], [0, gap_frame], [0, 0]]), 'constant')
# print("reshape aim_mel")
# else:
# print("这里！ori_mel_batch.shape[1] = " + str(ori_mel.shape[1]))
# ori_mel = np.pad(ori_mel, ([[0, 0], [0, gap_frame], [0, 0]]), 'constant')
# print("reshape ori_mel")
#
# print("2 ori_mel_batch.shape = " + str(ori_mel.shape[1]) + " aim_mel_batch.shape = " + str(aim_mel.shape[1]))
#
#
# except tf.errors.OutOfRangeError:
# print("complete")
# finally:
# coord.request_stop()
# coord.join(threads)
# sess.close()
"""
ori_mel_batch.shape = (16, 547, 80) aim_mel_batch.shape = (16, 743, 80)
ori_mel_batch.shape = (16, 890, 80) aim_mel_batch.shape = (16, 750, 80)
ori_mel_batch.shape = (16, 668, 80) aim_mel_batch.shape = (16, 663, 80)
ori_mel_batch.shape = (16, 737, 80) aim_mel_batch.shape = (16, 756, 80)
ori_mel_batch.shape = (16, 817, 80) aim_mel_batch.shape = (16, 743, 80)
ori_mel_batch.shape = (16, 565, 80) aim_mel_batch.shape = (16, 663, 80)
ori_mel_batch.shape = (16, 737, 80) aim_mel_batch.shape = (16, 750, 80)
ori_mel_batch.shape = (16, 890, 80) aim_mel_batch.shape = (16, 728, 80)
ori_mel_batch.shape = (16, 890, 80) aim_mel_batch.shape = (16, 750, 80)
ori_mel_batch.shape = 817 aim_mel_batch.shape = 663
817
ori_mel_batch.shape = 788 aim_mel_batch.shape = 756
788
ori_mel_batch.shape = 737 aim_mel_batch.shape = 648
737
ori_mel_batch.shape = 817 aim_mel_batch.shape = 750
817
ori_mel_batch.shape = 822 aim_mel_batch.shape = 756
法1：在train.py里面，使用同一个 batch 数据之前，把两个mel 数据pad对齐一下，再送进 G，生成模型（）
法2：在get_next_batch（）函数返回之前，处理好尺寸；
"""
"""
1 ori_mel_batch.shape = 890 aim_mel_batch.shape = 608
max_frame = 890 gap_frame = 282
reshape aim_mel
2 ori_mel_batch.shape = 890 aim_mel_batch.shape = 890
1 ori_mel_batch.shape = 802 aim_mel_batch.shape = 756
max_frame = 802 gap_frame = 46
reshape aim_mel
2 ori_mel_batch.shape = 802 aim_mel_batch.shape = 802
1 ori_mel_batch.shape = 817 aim_mel_batch.shape = 743
max_frame = 817 gap_frame = 74
reshape aim_mel
2 ori_mel_batch.shape = 817 aim_mel_batch.shape = 817
1 ori_mel_batch.shape = 788 aim_mel_batch.shape = 648
max_frame = 788 gap_frame = 140
reshape aim_mel
2 ori_mel_batch.shape = 788 aim_mel_batch.shape = 788
1 ori_mel_batch.shape = 822 aim_mel_batch.shape = 756
max_frame = 822 gap_frame = 66
reshape aim_mel
2 ori_mel_batch.shape = 822 aim_mel_batch.shape = 822
"""