frontend
acoustic_base.py
acoustic_composition.py
acoustic_normalisation.py
feature_normalisation_base.py
label_composer.py
label_modifier.py
label_normalisation.py
linguistic_base.py
mean_variance_norm.py
merge_features.py
min_max_norm.py
mlpg.py
mlpg_fast.py
parameter_generation.py
silence_remover.py
configuration
configuration.py
将用户的所有外部定义的配置文件解析成一个ConfigParser类
examplelabelconfigfile.py
DNN的所有输入lab特征的配置类
logplot
logging_plotting.py
models
deep_rnn.py
DeepRecurrentNetwork
dnn.py
DNN
dnn_cm.py
DNN
hed_rnn.py
DeepEncoderDecoderNetwork
mdn.py
MixtureDensityNetwork
seq2seq.py
VanillaSequenceEncoder
VanillaSequenceEncoderWithDur
DistributedSequenceEncoder
st_dnn_cm.py
SequentialDNN
layers
gating.py
VanillaRNN
标准RNN隐藏神经元
VanillaRNNDecoder
标准RNN解码器
LstmBase
LSTM基础类,相关函数
LstmDecoderBase
LSTM基础类,以及相关类
VanillaLstm
LSTM基础block
VanillaLstmDecoder
LSTM基础block
SimplifiedLstmDecoder
只保留遗忘门的LSTM block
LstmNFG
去掉遗忘门的LSTM block
LstmNIG
去掉输入门的LSTM block
LstmNOG
去掉输出门的LSTM block
LstmNoPeepholes
没有孔洞连接的LSTM block
SimplifiedLstm
只保留遗忘门的LSTM block
SimplifiedGRU
只保留遗忘门的GRU
BidirectionSLstm
基于SimplifiedLstm
BidirectionLstm
基于VanillaLstm
RecurrentOutput
GatedRecurrentUnit
layers.py
MixtureDensityOutputLayer
LinearLayer
SigmoidLayer
GeneralLayer
任意激活函数(默认为linear)的前馈层
HiddenLayer
SplitHiddenLayer
TokenProjectionLayer
dA
lhuc_layer.py
SigmoidLayer_LHUC
LstmBase_LHUC
VanillaLstm_LHUC
mdn_layers.py
MixtureDensityOutputLayer
LinearLayer
SigmoidLayer
GeneralLayer
HiddenLayer
dA
recurrent_decoders.py
VanillaRNNDecoder
ContextRNNDecoder
LstmDecoderBase
VanillaLstmDecoder
SimplifiedLstmDecoder
LstmBase
ContextLstm
recurrent_output_layer.py
RecurrentOutputLayer
training_schemes
adam.py
手动实现ADAM参数更新算法
adam_v2.py
手动实现ADAM参数更新算法
rprop.py
手动实现RPROP
io_funcs
binary_io.py
numpy 数组读取和存入二进制文件
htk_io.py
读取和写入htk格式的二进制文件
work_in_progress
utils
acous_feat_extraction.py
使用magphase抽取特征,python包的形式
compute_distortion.py
计算扭曲度,失真度等,计算 f0,lf0,mgc,bap等的mse,corr等
file_paths.py
根据配置文件中的参数,读取所有目录下的所有文件到列表
generate.py
假若当前不存在c语言版本的STRAIGHT,用来合成语音的脚本
learn_rates.py
调整学习率,动态减少
providers.py
为深度学习模型载入数据到CPU或GPU的,utterance by utterance 或block by block,类似tensorflow的batch_generator
utils.py
读取文件列表,或者返回文件夹下的文件列表
view.py
直接在屏幕上打印 一些二进制文件,用于调试
tensorflow_lib
configuration.py
data_utils.py
model.py
train.py
keras_lib
configuration.py
data_utils.py
model.py
train.py
validation.py
run_tensorflow_with_merlin_io.py
run_merlin_hed.py
run_keras_with_merlin_io.py
run_merlin.py
gpu_lock.py