EMSAssist

This repository currently contains the reproducible artifact for EMSAssist.

A. Preparing Docker Environment (for using Docker)

We follow the official Docker guide to install and run Docker images:

Assuming NVIDIA GPUs present in the bare metal system running Ubuntu 22.04.

Install Docker

Update and Install Docker:

$ sudo apt update
$ sudo apt-get install docker

Test Docker installation: (should show docker.service details.)

$ sudo systemctl status docker

Perform post installation steps to avoid sudo

# Create the Docker group.
$ sudo groupadd docker
# Add your user to the Docker group
$ sudo usermod -aG docker $USER
# Activate the changes to groups
$ newgrp docker

Build Docker image with all required packages. This step is optional as a ready image will be pulled in absence from Docker hub (step 7). This step involves manual installation of packages and rebuilding the image which will take some time.

Create Dockerfile

touch Dockerfile

Edit Dockerfile with your preffered text editor and paste the following contents in Dockerfile then save it.

#FROM ubuntu:22.04
FROM nvidia/cuda:12.0.1-base-ubuntu22.04
MAINTAINER "Amran Haroon"

ENV LANG=C.UTF-8 LC_ALL=C.UTF-8
ENV PATH /opt/conda/bin:$PATH

RUN set -x && \
	apt-get update --fix-missing && \
	apt-get install -y --no-install-recommends \
		bzip2 \
		ca-certificates \
		git \
		libglib2.0-0 \
		libsm6 \
		libxcomposite1 \
		libxcursor1 \
		libxdamage1 \
		libxext6 \
		libxfixes3 \
		libxi6 \
		libxinerama1 \
		libxrandr2 \
		libxrender1 \
		mercurial \
		openssh-client \
		procps \
		subversion \
		wget \
	&& apt-get clean \
	&& rm -rf /var/lib/apt/lists/* && \
	UNAME_M="$(uname -m)" && \
	if [ "${UNAME_M}" = "x86_64" ]; then \
		ANACONDA_URL="https://repo.anaconda.com/archive/Anaconda3-2023.03-Linux-x86_64.sh"; \
		SHA256SUM="19737d5c27b23a1d8740c5cb2414bf6253184ce745d0a912bb235a212a15e075"; \
	elif [ "${UNAME_M}" = "s390x" ]; then \
		ANACONDA_URL="https://repo.anaconda.com/archive/Anaconda3-2022.10-Linux-s390x.sh"; \
		SHA256SUM="f5ccc24aedab1f3f9cccf1945ca1061bee194fa42a212ec26425f3b77fdd943a"; \
	elif [ "${UNAME_M}" = "aarch64" ]; then \
		ANACONDA_URL="https://repo.anaconda.com/archive/Anaconda3-2022.10-Linux-aarch64.sh"; \
		SHA256SUM="fbadbfae5992a8c96af0a4621262080eea44e22baee2172e3dfb640f5cf8d22d"; \
	elif [ "${UNAME_M}" = "ppc64le" ]; then \
		ANACONDA_URL="https://repo.anaconda.com/archive/Anaconda3-2022.10-Linux-ppc64le.sh"; \
		SHA256SUM="8fdebc79f63b74daad421a2674d43299fa9c5007d85cf00e8dc1a81fbf2787e4"; \
	fi && \
	wget "${ANACONDA_URL}" -O anaconda.sh -q && \
	echo "${SHA256SUM} anaconda.sh" > shasum && \
	sha256sum --check --status shasum && \
	/bin/bash anaconda.sh -b -p /opt/conda && \
	rm anaconda.sh shasum && \
	ln -s /opt/conda/etc/profile.d/conda.sh /etc/profile.d/conda.sh && \
	echo ". /opt/conda/etc/profile.d/conda.sh" >> ~/.bashrc && \
	echo "conda activate base" >> ~/.bashrc && \
	find /opt/conda/ -follow -type f -name '*.a' -delete && \
	find /opt/conda/ -follow -type f -name '*.js.map' -delete && \
	/opt/conda/bin/conda clean -afy

RUN set -x && \     
	conda create -yn emsassist-gpu pip python=3.7 

#ENTRYPOINT ["tail", "-f", "/dev/null"]
#CMD [ "/bin/bash" ]
ENTRYPOINT ["tail"]
CMD ["-f","/dev/null"]

Build Docker image and give it a name:

$ docker build -t haroon3rd/anaconda3:base .

Make sure the image was created successfully

$ docker image ls

Run Docker image with gpu enabled in silent mode and execute into terminal :

$ docker run --gpus all -d -t --name base haroon3rd/anaconda3:base
$ docker exec -it base /bin/bash

Install the required packages inside created conda environment:

# Activate conda env 
$ conda activate emsassist-gpu

# Install gcc
$ apt-get update && apt-get -y install gcc mono-mcs && rm -rf /var/lib/apt/lists/*

# Install the required python modules one by one
$ pip install tensorflow_addons sentencepiece gin-config tflite-support tensorflow_hub natsort scikit-learn fire pyyaml tqdm librosa tensorflow-io==0.26 tensorflow_datasets nltk pydub pandas

# Install the XGBoost-GPU
$ conda install py-xgboost-gpu
# This also installs the CudaToolkit: pkgs/main/linux-64::cudatoolkit-10.0.130-0

# Install the TensorFlow-2.9
$ pip install tensorflow-gpu==2.9

# Install the CUDA ToolKit 11.0 and CuDNN 8.0
$ conda install -c conda-forge cudatoolkit=11.0 cudnn

Save this container (with changes) to a new image (may take up to 10 mins):

$ docker commit base  haroon3rd/anaconda3:nvidia-v1

Make sure you have a new image haroon3rd/anaconda3:nvidia-v1:

$ docker image ls

Install Docker-Compose

$ sudo apt update
$ sudo apt-get install docker-compose

install nvidia container tootlkit in local machine (requires Docker dameon to reload). This toolkit allows you to connect the container engine to the bare metal machine's nvidia driver.

# Add key and repository 
$ curl -s -L https://nvidia.github.io/nvidia-docker/gpgkey | apt-key add -
$ curl -s -L https://nvidia.github.io/nvidia-docker/ubuntu22.04/nvidia-docker.list > /etc/apt/sources.list.d/nvidia-docker.list

# Update repository
$ apt update

# Install nvidia-container-toolkit
$ sudo apt -y install nvidia-container-toolkit

# Restart docker engine
$ systemctl restart docker

Download the data.tar.gz, model.tar.gz and docker-compose.yml files from Google Drive to the cuurent working (i.e., ./home) folder. We expect the downloading and decompressing to take 2-3 hours.

decompress the model.tar.gz: tar -xvzf model.tar.gz
decompress the data.tar.gz: tar -xvzf data.tar.gz. After this step, make sure we have 4 items in the current directory: data, model, and EMSAssist folder and a file 'docker-compose.yml'. Also make sure there are 3 folders under EMSAssist directory: src, examples, and init_models.
Make sure docker-compose.yml has the following content in it:

version: '3.7'

services:
  emsassist:
    image: haroon3rd/anaconda3:nvidia-v1
      container_name: emsassist
      volumes:
        - ./data:/home/EMSAssist-artifact-evaluation/data
        - ./model:/home/EMSAssist-artifact-evaluation/model
        - ./EMSAssist:/home/EMSAssist-artifact-evaluation/EMSAssist
      #command: [/bin/bash -c "tail -f /dev/null"]
      command: tail -F anything
      #network_mode: "host"
      deploy:
        resources:
          reservations:
            devices:
              - driver: nvidia
                capabilities: [gpu]
volumes:
  emsassist: {}

Clone the git repository of EMSAssist:

$ git clone --recursive git@github.com:LENSS/EMSAssist.git`
$ cd EMSAssist
$ git clone --recursive git@github.com:tensorflow/examples.git

Run docker-compose in silent mode from the terminal of current folder. For the next step, data, model, and EMSAssist directories along with the docker-compose.yml file need to be in the same folder (i.e., current folder):
```
$ docker-compose up -d
# it will pull a docker container image and run it in local machine as "emsassist"
```
Log in to your running container's terminal (bash)
```
$ docker exec -it emsassist /bin/bash
```

Inside your container:

Activate conda environment and run nvidia-smi to make sure the GPU is working:
```
$ conda activate emsassist-gpu
$ nvidia-smi
```

B. Path preparation and running the evaluation

Before we proceed, please make sure you successfully set up the environment or get the Docker image running with nvidia-smi

 $ export PYTHONPATH=$PWD/src/speech_recognition:$PWD/examples`

 $ export LD_LIBRARY_PATH=$CONDA_PREFIX/lib:$LD_LIBRARY_PATH`

 # begin the evaluation: 
 $ cd src
 # Check out the README file when you enter each sub-directory.

Basic Environment

Software Environment	Version
OS	Ubuntu Server 22.04.1 LTS
NVIDIA Driver	525.85.12

Hardware Environment	Version
GPU	3 x NVIDIA A30
CPU	2 x Intel Xeon 4314

Before the artifact evaluation and use the open-sourced code/data, please make sure you have at least 1 NVIDIA GPU available with nvidia-smi command.

Build the target Environment

Software Environment	Version
OS	Ubuntu Server 22.04.1 LTS
NVIDIA Driver	525.85.12
CUDA Version	10
TensorFlow	2.9

haroon3rd / readme

EMSAssist

A. Preparing Docker Environment (for using Docker)

B. Path preparation and running the evaluation

Basic Environment

Build the target Environment

About