1. frontend
    1. acoustic_base.py
    2. acoustic_composition.py
    3. acoustic_normalisation.py
    4. feature_normalisation_base.py
    5. label_composer.py
    6. label_modifier.py
    7. label_normalisation.py
    8. linguistic_base.py
    9. mean_variance_norm.py
    10. merge_features.py
    11. min_max_norm.py
    12. mlpg.py
    13. mlpg_fast.py
    14. parameter_generation.py
    15. silence_remover.py
  2. configuration
    1. configuration.py
      1. 将用户的所有外部定义的配置文件解析成一个ConfigParser类
    2. examplelabelconfigfile.py
      1. DNN的所有输入lab特征的配置类
  3. logplot
    1. logging_plotting.py
  4. models
    1. deep_rnn.py
      1. DeepRecurrentNetwork
    2. dnn.py
      1. DNN
    3. dnn_cm.py
      1. DNN
    4. hed_rnn.py
      1. DeepEncoderDecoderNetwork
    5. mdn.py
      1. MixtureDensityNetwork
    6. seq2seq.py
      1. VanillaSequenceEncoder
      2. VanillaSequenceEncoderWithDur
      3. DistributedSequenceEncoder
    7. st_dnn_cm.py
      1. SequentialDNN
  5. layers
    1. gating.py
      1. VanillaRNN
        1. 标准RNN隐藏神经元
      2. VanillaRNNDecoder
        1. 标准RNN解码器
      3. LstmBase
        1. LSTM基础类,相关函数
      4. LstmDecoderBase
        1. LSTM基础类,以及相关类
      5. VanillaLstm
        1. LSTM基础block
      6. VanillaLstmDecoder
        1. LSTM基础block
      7. SimplifiedLstmDecoder
        1. 只保留遗忘门的LSTM block
      8. LstmNFG
        1. 去掉遗忘门的LSTM block
      9. LstmNIG
        1. 去掉输入门的LSTM block
      10. LstmNOG
        1. 去掉输出门的LSTM block
      11. LstmNoPeepholes
        1. 没有孔洞连接的LSTM block
      12. SimplifiedLstm
        1. 只保留遗忘门的LSTM block
      13. SimplifiedGRU
        1. 只保留遗忘门的GRU
      14. BidirectionSLstm
        1. 基于SimplifiedLstm
      15. BidirectionLstm
        1. 基于VanillaLstm
      16. RecurrentOutput
      17. GatedRecurrentUnit
    2. layers.py
      1. MixtureDensityOutputLayer
      2. LinearLayer
      3. SigmoidLayer
      4. GeneralLayer
        1. 任意激活函数(默认为linear)的前馈层
      5. HiddenLayer
      6. SplitHiddenLayer
      7. TokenProjectionLayer
      8. dA
    3. lhuc_layer.py
      1. SigmoidLayer_LHUC
      2. LstmBase_LHUC
      3. VanillaLstm_LHUC
    4. mdn_layers.py
      1. MixtureDensityOutputLayer
      2. LinearLayer
      3. SigmoidLayer
      4. GeneralLayer
      5. HiddenLayer
      6. dA
    5. recurrent_decoders.py
      1. VanillaRNNDecoder
      2. ContextRNNDecoder
      3. LstmDecoderBase
      4. VanillaLstmDecoder
      5. SimplifiedLstmDecoder
      6. LstmBase
      7. ContextLstm
    6. recurrent_output_layer.py
      1. RecurrentOutputLayer
  6. training_schemes
    1. adam.py
      1. 手动实现ADAM参数更新算法
    2. adam_v2.py
      1. 手动实现ADAM参数更新算法
    3. rprop.py
      1. 手动实现RPROP
  7. io_funcs
    1. binary_io.py
      1. numpy 数组读取和存入二进制文件
    2. htk_io.py
      1. 读取和写入htk格式的二进制文件
  8. work_in_progress
  9. utils
    1. acous_feat_extraction.py
      1. 使用magphase抽取特征,python包的形式
    2. compute_distortion.py
      1. 计算扭曲度,失真度等,计算 f0,lf0,mgc,bap等的mse,corr等
    3. file_paths.py
      1. 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
    4. generate.py
      1. 假若当前不存在c语言版本的STRAIGHT,用来合成语音的脚本
    5. learn_rates.py
      1. 调整学习率,动态减少
    6. providers.py
      1. 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
    7. utils.py
      1. 读取文件列表,或者返回文件夹下的文件列表
    8. view.py
      1. 直接在屏幕上打印 一些二进制文件,用于调试
  10. tensorflow_lib
    1. configuration.py
    2. data_utils.py
    3. model.py
    4. train.py
  11. keras_lib
    1. configuration.py
    2. data_utils.py
    3. model.py
    4. train.py
  12. validation.py
  13. run_tensorflow_with_merlin_io.py
  14. run_merlin_hed.py
  15. run_keras_with_merlin_io.py
  16. run_merlin.py
  17. gpu_lock.py