Kaldi is a very popular Automatic Speech Recognition (ASR) toolkit. In my experience, setting up Kaldi on a system can take huge amount of time. Therefore, it is best to use containerization approach.
Following is the Dockerfile that I wrote. We will be using a cuda enabled Ubuntu 18.04 image from Nvidia NGC:
FROM nvidia/cuda:11.1.1-cudnn8-devel-ubuntu18.04
LABEL Author="umerjamil16@github.io" Email="umer1694@gmail.com"
LABEL Description="Kaldi ASR Image" Vendor="x" Version="1.0"
RUN sh -c 'echo "APT { Get { AllowUnauthenticated \"1\"; }; };" > /etc/apt/apt.conf.d/99allow_unauth'
RUN apt -o Acquire::AllowInsecureRepositories=true -o Acquire::AllowDowngradeToInsecureRepositories=true update
RUN apt-get install -y curl wget
RUN apt-key del 7fa2af80
RUN wget https://developer.download.nvidia.com/compute/cuda/repos/ubuntu2004/x86_64/cuda-keyring_1.0-1_all.deb
RUN dpkg -i cuda-keyring_1.0-1_all.deb
RUN rm -f /etc/apt/sources.list.d/cuda.list /etc/apt/apt.conf.d/99allow_unauth cuda-keyring_1.0-1_all.deb
RUN apt-key adv --keyserver keyserver.ubuntu.com --recv-keys A4B469963BF863CC F60F4B3D7FA2AF80
RUN apt-get update && apt-get upgrade -y
RUN apt-get install git -y
WORKDIR /opt
RUN git clone --depth 1 https://github.com/kaldi-asr/kaldi.git
WORKDIR /opt/kaldi/tools
RUN apt-get install zlib1g-dev automake autoconf unzip wget git sox gfortran libtool subversion python2.7 python3 flac gawk swig python-pip nano -y
RUN ./extras/install_mkl.sh
RUN pip install numpy
#RUN pip install git+https://github.com/sequitur-g2p/sequitur-g2p@master
RUN make -j $(nproc)
#RUN ./extras/install_irstlm.sh
RUN ./extras/install_srilm.sh name org email
RUN ./extras/install_sequitur.sh
RUN chmod +x env.sh
RUN ./env.sh
WORKDIR /opt/kaldi/src/
RUN ./configure --shared --use-cuda
RUN make depend -j $(nproc)
RUN make -j $(nproc)
RUN echo "SUCCESS"
WORKDIR /opt/kaldi/
Building the image
To build the docker image, cd to folder where Kaldi Dockerfile is located and run the following command.
docker build -t kaldi_asr:latest -f Dockerfile .
This will build the image with the kaldi_asr name.
Running the kaldi container
First let us create a working_directory to mount it to our container. The working_dir will contain our Kaldi project files and will allow data persist.
mkdir working_dir
docker run -it --rm -v ./working_dir:/working_dir --name kaldi_docker kaldi_asr:latest
To attach bash shell to the above running container,
docker exec -it kaldi_docker /bin/bash
Now you can run your kaldi scripts inside this container and work on your projects.