GithubHelp home page GithubHelp logo

rnn-transducer-1's Introduction

RNN-Transducer

  • 模型训练过程
  1. 网络结构 网络结构主要采取多层LSTM,encoder使用五层,decoder使用两层,每层之间加入一层非线性变换Relu。 #参数设计 encoder: hidden_size:1024 project_size:300 layer_norm

    decoder: hidden_size:512 project_size:300 layer_norm

    jointNet: hidden_size:2048

  2. 预训练方法

    1. 声学模型采用CTC进行预训练。原始paper里采用层级CTC预训练。从多个维度对语音特征进行关系映射。这里直接对中文word进行预训练即可。得到语音特征到文字的映射关系。
    2. 语言模型采用LSTM进行预训练即可。从前文对下个节点进行结果预测。
  3. 模型细节:

  • 帧移25ms,帧长10ms
  • 输入:40维度的Fbank特征,[B,L,40],L最大设置为1600
  • Low Frame Rate(10, 5), [B,L//5, 400], 对语音帧进行合并,合并10帧,然后跳过5帧,继续合并10帧。
  • Encoder: 5 * (LN + LSTM + projection_layer) = [B, 320, 400] 输出:[B, 320, 320]
  • Decoder: 2 * (LN + LSTM + projection_layer) = [B, 50, 512] 输出:[B, 50, 320]
  • Joint: concat + forward(2048) + tanh + RNNT_loss解码
  1. 模型优化点:

    1. load_randomly_augmented_audio,语音增益 与 语速增益
    2. encoder预训练采用原始数据,收敛后,再添加40%的specAugment(from google)扩增数据,直到收敛。
    3. 收敛后,融合decoder使用带specAugment扩增的数据进行训练。
  2. 三个模块的参数组合设置

    1. 三个模块不同参数组合对是否能收敛影响很大。
    2. 编码器数量最少要是解码器参数的四倍以上。
    3. JointNetWork参数比解码器参数要大。
    4. 需要先将编码器encoder loss训练到最优地步,loss<0.1
    5. decoder训练到loss<2
    6. 联合encoder,decoder,进行训练

实验结果: rnnt 第一版 wer=12.31

rnn-transducer-1's People

Recommend Projects

  • React photo React

    A declarative, efficient, and flexible JavaScript library for building user interfaces.

  • Vue.js photo Vue.js

    🖖 Vue.js is a progressive, incrementally-adoptable JavaScript framework for building UI on the web.

  • Typescript photo Typescript

    TypeScript is a superset of JavaScript that compiles to clean JavaScript output.

  • TensorFlow photo TensorFlow

    An Open Source Machine Learning Framework for Everyone

  • Django photo Django

    The Web framework for perfectionists with deadlines.

  • D3 photo D3

    Bring data to life with SVG, Canvas and HTML. 📊📈🎉

Recommend Topics

  • javascript

    JavaScript (JS) is a lightweight interpreted programming language with first-class functions.

  • web

    Some thing interesting about web. New door for the world.

  • server

    A server is a program made to process requests and deliver data to clients.

  • Machine learning

    Machine learning is a way of modeling and interpreting data that allows a piece of software to respond intelligently.

  • Game

    Some thing interesting about game, make everyone happy.

Recommend Org

  • Facebook photo Facebook

    We are working to build community through open source technology. NB: members must have two-factor auth.

  • Microsoft photo Microsoft

    Open source projects and samples from Microsoft.

  • Google photo Google

    Google ❤️ Open Source for everyone.

  • D3 photo D3

    Data-Driven Documents codes.