PIKA: a lightweight speech processing toolkit based on Pytorch and (Py)Kaldi

Related tags

Deep Learningpika
Overview

PIKA: a lightweight speech processing toolkit based on Pytorch and (Py)Kaldi

PIKA is a lightweight speech processing toolkit based on Pytorch and (Py)Kaldi. The first release focuses on end-to-end speech recognition. We use Pytorch as deep learning engine, Kaldi for data formatting and feature extraction.

Key Features

  • On-the-fly data augmentation and feature extraction loader

  • TDNN Transformer encoder and convolution and transformer based decoder model structure

  • RNNT training and batch decoding

  • RNNT decoding with external Ngram FSTs (on-the-fly rescoring, aka, shallow fusion)

  • RNNT Minimum Bayes Risk (MBR) training

  • LAS forward and backward rescorer for RNNT

  • Efficient BMUF (Block model update filtering) based distributed training

Installation and Dependencies

In general, we recommend Anaconda since it comes with most dependencies. Other major dependencies include,

Pytorch

Please go to https://pytorch.org/ for pytorch installation, codes and scripts should be able to run against pytorch 0.4.0 and above. But we recommend 1.0.0 above for compatibility with RNNT loss module (see below)

Pykaldi and Kaldi

We use Kaldi (https://github.com/kaldi-asr/kaldi)) and PyKaldi (a python wrapper for Kaldi) for data processing, feature extraction and FST manipulations. Please go to Pykaldi website https://github.com/pykaldi/pykaldi for installation and make sure to build Pykaldi with ninja for efficiency. After following the installation process of pykaldi, you should have both Kaldi and Pykaldi dependencies ready.

CUDA-Warp RNN-Transducer

For RNNT loss module, we adopt the pytorch binding at https://github.com/1ytic/warp-rnnt

Others

Check requirements.txt for other dependencies.

Get Started

To get started, check all the training and decoding scripts located in egs directory.

I. Data preparation and RNNT training

egs/train_transducer_bmuf_otfaug.sh contains data preparation and RNNT training. One need to prepare training data and specify the training data directory,

#training data dir must contain wav.scp and label.txt files
#wav.scp: standard kaldi wav.scp file, see https://kaldi-asr.org/doc/data_prep.html 
#label.txt: label text file, the format is, uttid sequence-of-integer, where integer
#           is one-based indexing mapped label, note that zero is reserved for blank,  
#           ,eg., utt_id_1 3 5 7 10 23 
train_data_dir=

II. Continue with MBR training

With RNNT trained model, one can continued MBR training with egs/train_transducer_mbr_bmuf_otfaug.sh (assuming using the same training data, therefore data preparation is omitted). Make sure to specify the initial model,

--verbose \
--optim sgd \
--init_model $exp_dir/init.model \
--rnnt_scale 1.0 \
--sm_scale 0.8 \

III. Training LAS forward and backward rescorer

One can train a forward and backward LAS rescorer for your RNN-T model using egs/train_las_rescorer_bmuf_otfaug.sh. The LAS rescorer will share the encoder part with RNNT model, and has extra two-layer LSTM as additional encoder, make sure to specify the encoder sharing as,

--num_batches_per_epoch 526264 \
--shared_encoder_model $exp_dir/final.model \
--num_epochs 5 \

We support bi-directional LAS rescoring, i.e., forward and backward rescoring. Backward (right-to-left) rescoring is achieved by reversing sequential labels when conducting LAS model training. One can easily perform a backward LAS rescorer training by specifying,

--reverse_labels

IV. Decoding

egs/eval_transducer.sh is the main evluation script, which contains the decoding pipeline. Forward and backward LAS rescoring can be enabled by specifying these two models,

##########configs#############
#rnn transducer model
rnnt_model=
#forward and backward las rescorer model
lasrescorer_fw=
lasrescorer_bw=

Caveats

All the training and decoding hyper-parameters are adopted based on large-scale (e.g., 60khrs) training and internal evaluation data. One might need to re-tune hyper-parameters to acheive optimal performances. Also the WER (CER) scoring script is based on a Mandarin task, we recommend those who work on different languages rewrite scoring scripts.

References

[1] Improving Attention Based Sequence-to-Sequence Models for End-to-End English Conversational Speech Recognition, Chao Weng, Jia Cui, Guangsen Wang, Jun Wang, Chengzhu Yu, Dan Su, Dong Yu, InterSpeech 2018

[2] Minimum Bayes Risk Training of RNN-Transducer for End-to-End Speech Recognition, Chao Weng, Chengzhu Yu, Jia Cui, Chunlei Zhang, Dong Yu, InterSpeech 2020

Citations

@inproceedings{Weng2020,
  author={Chao Weng and Chengzhu Yu and Jia Cui and Chunlei Zhang and Dong Yu},
  title={{Minimum Bayes Risk Training of RNN-Transducer for End-to-End Speech Recognition}},
  year=2020,
  booktitle={Proc. Interspeech 2020},
  pages={966--970},
  doi={10.21437/Interspeech.2020-1221},
  url={http://dx.doi.org/10.21437/Interspeech.2020-1221}
}

@inproceedings{Weng2018,
  author={Chao Weng and Jia Cui and Guangsen Wang and Jun Wang and Chengzhu Yu and Dan Su and Dong Yu},
  title={Improving Attention Based Sequence-to-Sequence Models for End-to-End English Conversational Speech Recognition},
  year=2018,
  booktitle={Proc. Interspeech 2018},
  pages={761--765},
  doi={10.21437/Interspeech.2018-1030},
  url={http://dx.doi.org/10.21437/Interspeech.2018-1030}
}

Disclaimer

This is not an officially supported Tencent product

Owner
Research repositories.
Consensus Learning from Heterogeneous Objectives for One-Class Collaborative Filtering

Consensus Learning from Heterogeneous Objectives for One-Class Collaborative Filtering This repository provides the source code of "Consensus Learning

SeongKu-Kang 6 Apr 29, 2022
3DIAS: 3D Shape Reconstruction with Implicit Algebraic Surfaces (ICCV 2021)

3DIAS_Pytorch This repository contains the official code to reproduce the results from the paper: 3DIAS: 3D Shape Reconstruction with Implicit Algebra

Mohsen Yavartanoo 21 Dec 12, 2022
DimReductionClustering - Dimensionality Reduction + Clustering + Unsupervised Score Metrics

Dimensionality Reduction + Clustering + Unsupervised Score Metrics Introduction

11 Nov 15, 2022
🥇Samsung AI Challenge 2021 1등 솔루션입니다🥇

MoT - Molecular Transformer Large-scale Pretraining for Molecular Property Prediction Samsung AI Challenge for Scientific Discovery This repository is

Jungwoo Park 44 Dec 03, 2022
Video Representation Learning by Recognizing Temporal Transformations. In ECCV, 2020.

Video Representation Learning by Recognizing Temporal Transformations [Project Page] Simon Jenni, Givi Meishvili, and Paolo Favaro. In ECCV, 2020. Thi

Simon Jenni 46 Nov 14, 2022
An End-to-End Machine Learning Library to Optimize AUC (AUROC, AUPRC).

Logo by Zhuoning Yuan LibAUC: A Machine Learning Library for AUC Optimization Website | Updates | Installation | Tutorial | Research | Github LibAUC a

Optimization for AI 176 Jan 07, 2023
Implementation for paper: Self-Regulation for Semantic Segmentation

Self-Regulation for Semantic Segmentation This is the PyTorch implementation for paper Self-Regulation for Semantic Segmentation, ICCV 2021. Citing SR

Dong ZHANG 30 Nov 21, 2022
Official implementation of "Can You Spot the Chameleon? Adversarially Camouflaging Images from Co-Salient Object Detection" in CVPR 2022.

Jadena Official implementation of "Can You Spot the Chameleon? Adversarially Camouflaging Images from Co-Salient Object Detection" in CVPR 2022. arXiv

Qing Guo 13 Nov 29, 2022
Pytorch library for seismic data augmentation

Pytorch library for seismic data augmentation

Artemii Novoselov 27 Nov 22, 2022
A Multi-modal Model Chinese Spell Checker Released on ACL2021.

ReaLiSe ReaLiSe is a multi-modal Chinese spell checking model. This the office code for the paper Read, Listen, and See: Leveraging Multimodal Informa

DaDa 106 Dec 29, 2022
This repository contains the reference implementation for our proposed Convolutional CRFs.

ConvCRF This repository contains the reference implementation for our proposed Convolutional CRFs in PyTorch (Tensorflow planned). The two main entry-

Marvin Teichmann 553 Dec 07, 2022
Must-read Papers on Physics-Informed Neural Networks.

PINNpapers Contributed by IDRL lab. Introduction Physics-Informed Neural Network (PINN) has achieved great success in scientific computing since 2017.

IDRL 330 Jan 07, 2023
Memory-efficient optimum einsum using opt_einsum planning and PyTorch kernels.

opt-einsum-torch There have been many implementations of Einstein's summation. numpy's numpy.einsum is the least efficient one as it only runs in sing

Haoyan Huo 9 Nov 18, 2022
library for nonlinear optimization, wrapping many algorithms for global and local, constrained or unconstrained, optimization

NLopt is a library for nonlinear local and global optimization, for functions with and without gradient information. It is designed as a simple, unifi

Steven G. Johnson 1.4k Dec 25, 2022
Official PyTorch Implementation of Convolutional Hough Matching Networks, CVPR 2021 (oral)

Convolutional Hough Matching Networks This is the implementation of the paper "Convolutional Hough Matching Network" by J. Min and M. Cho. Implemented

Juhong Min 70 Nov 22, 2022
sktime companion package for deep learning based on TensorFlow

NOTE: sktime-dl is currently being updated to work correctly with sktime 0.6, and wwill be fully relaunched over the summer. The plan is Refactor and

sktime 573 Jan 05, 2023
Beyond a Gaussian Denoiser: Residual Learning of Deep CNN for Image Denoising

Beyond a Gaussian Denoiser: Residual Learning of Deep CNN for Image Denoising

Kai Zhang 1.2k Dec 29, 2022
ONNX Runtime Web demo is an interactive demo portal showing real use cases running ONNX Runtime Web in VueJS.

ONNX Runtime Web demo is an interactive demo portal showing real use cases running ONNX Runtime Web in VueJS. It currently supports four examples for you to quickly experience the power of ONNX Runti

Microsoft 58 Dec 18, 2022
PyTorch implementation of Wide Residual Networks with 1-bit weights by McDonnell (ICLR 2018)

1-bit Wide ResNet PyTorch implementation of training 1-bit Wide ResNets from this paper: Training wide residual networks for deployment using a single

Sergey Zagoruyko 122 Dec 07, 2022
Code for our paper "Multi-scale Guided Attention for Medical Image Segmentation"

Medical Image Segmentation with Guided Attention This repository contains the code of our paper: "'Multi-scale self-guided attention for medical image

Ashish Sinha 394 Dec 28, 2022