eden-emu/opus
15
0
Fork 0
mirror of https://github.com/xiph/opus.git synced 2025-05-21 10:58:30 +00:00
Code Issues Projects Releases Packages Wiki Activity Actions 4

deeper features

Browse source
This commit is contained in:
Jean-Marc Valin 2018-10-03 22:30:44 -04:00
parent 639766b322
commit ea1391e174
3 changed files with 13 additions and 7 deletions
Download patch file Download diff file Expand all files Collapse all files

12
dnn/lpcnet.py
View file

@ -2,7 +2,7 @@
import math
from keras.models import Model
from keras.layers import Input, LSTM, CuDNNGRU, Dense, Embedding, Reshape, Concatenate, Lambda, Conv1D, Multiply, Bidirectional, MaxPooling1D, Activation
from keras.layers import Input, LSTM, CuDNNGRU, Dense, Embedding, Reshape, Concatenate, Lambda, Conv1D, Multiply, Add, Bidirectional, MaxPooling1D, Activation
from keras import backend as K
from keras.initializers import Initializer
from mdense import MDense
@ -47,14 +47,14 @@ def new_wavernn_model():
pitch = Input(shape=(None, 1))
feat = Input(shape=(None, nb_used_features))
pitch = Input(shape=(None, 1))
dec_feat = Input(shape=(None, 32))
dec_feat = Input(shape=(None, 128))
dec_state = Input(shape=(rnn_units,))
conv1 = Conv1D(16, 7, padding='causal', activation='tanh')
pconv1 = Conv1D(16, 5, padding='same', activation='tanh')
pconv2 = Conv1D(16, 5, padding='same', activation='tanh')
fconv1 = Conv1D(128, 3, padding='same', activation='tanh')
fconv2 = Conv1D(32, 3, padding='same', activation='tanh')
fconv2 = Conv1D(102, 3, padding='same', activation='tanh')
if False:
cpcm = conv1(pcm)
@ -73,6 +73,12 @@ def new_wavernn_model():
cfeat = fconv2(fconv1(cat_feat))
fdense1 = Dense(128, activation='tanh')
fdense2 = Dense(128, activation='tanh')
cfeat = Add()([cfeat, cat_feat])
cfeat = fdense2(fdense1(cfeat))
rep = Lambda(lambda x: K.repeat_elements(x, 160, 1))
rnn = CuDNNGRU(rnn_units, return_sequences=True, return_state=True)

4
dnn/test_wavenet_audio.py
View file

@ -66,7 +66,7 @@ in_data = np.reshape(in_data, (nb_frames*pcm_chunk_size, 1))
out_data = np.reshape(data, (nb_frames*pcm_chunk_size, 1))
model.load_weights('wavenet4d2_203.h5')
model.load_weights('wavenet4f3_30.h5')
order = 16
@ -75,7 +75,7 @@ fexc = np.zeros((1, 1, 2), dtype='float32')
iexc = np.zeros((1, 1, 1), dtype='int16')
state = np.zeros((1, lpcnet.rnn_units), dtype='float32')
for c in range(1, nb_frames):
cfeat = enc.predict(features[c:c+1, :, :nb_used_features])
cfeat = enc.predict([features[c:c+1, :, :nb_used_features], periods[c:c+1, :, :]])
for fr in range(1, feature_chunk_size):
f = c*feature_chunk_size + fr
a = features[c, fr, nb_features-order:]

4
dnn/train_wavenet_audio.py
View file

@ -104,8 +104,8 @@ in_data = np.concatenate([in_data, pred], axis=-1)
# f.create_dataset('data', data=in_data[:50000, :, :])
# f.create_dataset('feat', data=features[:50000, :, :])
checkpoint = ModelCheckpoint('wavenet4e_{epoch:02d}.h5')
checkpoint = ModelCheckpoint('wavenet4f3_{epoch:02d}.h5')
#model.load_weights('wavernn1c_01.h5')
#model.load_weights('wavenet4f2_30.h5')
model.compile(optimizer=Adam(0.001, amsgrad=True, decay=2e-4), loss='sparse_categorical_crossentropy', metrics=['sparse_categorical_accuracy'])
model.fit([in_data, in_exc, features, periods], out_data, batch_size=batch_size, epochs=30, validation_split=0.2, callbacks=[checkpoint])