eden-emu/opus
15
0
Fork 0
mirror of https://github.com/xiph/opus.git synced 2025-06-01 08:07:41 +00:00
Code Issues Projects Releases Packages Wiki Activity Actions

using features (except pitch gain which has NaNs for now)

Browse source
This commit is contained in:
Jean-Marc Valin 2018-06-26 01:31:44 -04:00
parent b65031ef64
commit 617e462be3
2 changed files with 12 additions and 5 deletions
Download patch file Download diff file Expand all files Collapse all files

11
dnn/lpcnet.py
View file

@ -12,12 +12,19 @@ import sys
rnn_units=64 rnn_units=64
pcm_bits = 8 pcm_bits = 8
pcm_levels = 2**pcm_bits pcm_levels = 2**pcm_bits
nb_used_features = 37
def new_wavernn_model(): def new_wavernn_model():
pcm = Input(shape=(None, 1)) pcm = Input(shape=(None, 1))
feat = Input(shape=(None, nb_used_features))
rep = Lambda(lambda x: K.repeat_elements(x, 160, 1))
rnn = CuDNNGRU(rnn_units, return_sequences=True) rnn = CuDNNGRU(rnn_units, return_sequences=True)
rnn_in = Concatenate()([pcm, rep(feat)])
md = MDense(pcm_levels, activation='softmax') md = MDense(pcm_levels, activation='softmax')
ulaw_prob = md(rnn(pcm)) ulaw_prob = md(rnn(rnn_in))
model = Model(pcm, ulaw_prob) model = Model([pcm, feat], ulaw_prob)
return model return model

6
dnn/train_lpcnet.py
View file

@ -24,7 +24,7 @@ model.summary()
pcmfile = sys.argv[1] pcmfile = sys.argv[1]
feature_file = sys.argv[2] feature_file = sys.argv[2]
nb_features = 54 nb_features = 54
nb_used_features = 38 nb_used_features = lpcnet.nb_used_features
feature_chunk_size = 15 feature_chunk_size = 15
pcm_chunk_size = 160*feature_chunk_size pcm_chunk_size = 160*feature_chunk_size
@ -44,8 +44,8 @@ out_data = (out_data.astype('int16')+128).astype('uint8')
features = np.reshape(features, (nb_frames, feature_chunk_size, nb_features)) features = np.reshape(features, (nb_frames, feature_chunk_size, nb_features))
features = features[:, :, :nb_used_features] features = features[:, :, :nb_used_features]
checkpoint = ModelCheckpoint('lpcnet1b_{epoch:02d}.h5') checkpoint = ModelCheckpoint('lpcnet1c_{epoch:02d}.h5')
#model.load_weights('wavernn1c_01.h5') #model.load_weights('wavernn1c_01.h5')
model.compile(optimizer=Adam(0.002, amsgrad=True, decay=2e-4), loss='sparse_categorical_crossentropy', metrics=['sparse_categorical_accuracy']) model.compile(optimizer=Adam(0.002, amsgrad=True, decay=2e-4), loss='sparse_categorical_crossentropy', metrics=['sparse_categorical_accuracy'])
model.fit(in_data, out_data, batch_size=batch_size, epochs=30, validation_split=0.2, callbacks=[checkpoint]) model.fit([in_data, features], out_data, batch_size=batch_size, epochs=30, validation_split=0.2, callbacks=[checkpoint])