Laptop to Container Examples

This project contains a set of example python codes that can be used with the laptop-to-container bash scripts for containerizing existing python projects.

Overview

These examples can be containerized using the Laptop to Container bash scripts without having to write your own Dockerfile.

Prerequisites

• laptop-to-container repository with bash scripts

Running the Scripts

Building with Conda

./create_docker.sh -a example-1 -d l2c/example-1 -e environment.yml -p linux/arm64 -m example1.py

./create_docker.sh -a example-2 -d l2c/example-2 -e environment.yml -m example2.py

./create_docker.sh -a example-3 -d l2c/example-3 -e environment.yml -m example3.py

./create_docker.sh -a example-4 -d l2c/example-4 -e environment.yml -m example4.py

./create_docker.sh -a example-5 -d l2c/example-5 -e environment.yml -m neural_style/neural_style.py

Parameter	Description	Default
-a	Path to the source code folder	N/A
-c	CUDA version to use	11.8
-d	Tag for generated docker image	latest
-e	Name of the environment yaml file in the source code folder	N/A
-h	Print help	N/A
-m	Python file containing the main method	N/A
-p	Build platform	linux/amd64
-r	Registry to publish the container image	None

To run the image, do the following:

docker run -it l2c/example-1

Building with Pip/CUDA

./create_docker_cuda.sh -a example-8 -d l2c/example-8 -e requirements.txt -m example8.py -p linux/arm64 -c 11.8

./create_docker_cuda.sh -a example-9 -d l2c/example-9 -e requirements.txt -m neural_style/neural_style.p -c 11.8

Parameter	Description	Default
-a	Path to the source code folder	N/A
-c	CUDA version to use	11.8
-d	Tag for generated docker image	latest
-e	Name of the pip environment file in the source code folder (e.g. requirements.txt)	N/A
-h	Print help	N/A
-m	Python file containing the main method	N/A
-p	Build platform	linux/amd64
-r	Registry to publish the container image	None

To run the image, do the following:

docker run -it l2c/example-8

Building with Pip

This is a ML example using LinearRegression that doesn't need CUDA.

./create_docker_pip.sh -a example-6 -d l2c/example-6 -e requirements.txt -p linux/arm64 -m example6.py

./create_docker_pip.sh -a example-7 -d l2c/example-7 -e requirements.txt -m example7.py

Parameter	Description	Default
-a	Path to the source code folder	N/A
-d	Tag for generated docker image	latest
-e	Name of the pip environment file in the source code folder (e.g. requirements.txt)	N/A
-h	Print help	N/A
-m	Python file containing the main method	N/A
-p	Build platform	linux/amd64
-r	Registry to publish the container image	None

To run the image, do the following:

docker run -it l2c/example-6

Running on Campus Cluster, Delta or DeltaAI

If you are pulling from an external registry, then you will need to login to the container registry first. This will create a local token.

apptainer registry login --username <username> docker://ghcr.io

./run_apptainer.sh -a <account> -p cpu -w /app -i l2c-example-1 -r ghcr.io/navarroc

./run_apptainer.sh -a <account> -p cpu -w /app -i l2c-example-2 -r ghcr.io/navarroc

./run_apptainer.sh -a <account> -p cpu -w /app -b "--bind /path/to/output:/app/outputs" -i l2c-example-5 -r ghcr.io/navarroc eval --content-image images/content-images/amber.jpg --model saved_models/mosaic.pth --output-image /app/outputs/output-normal2.jpg

In the above example, /path/to should be replaced with the path on your account where outputs should be written so the are available outside the container.

For DeltaAI, you will need to specify 1 or more GPUs or the job will not run.

Parameter	Description	Default
-a	Account to submit job under	N/A
-b	Bindings to mount when running the image	N/A
-d	Tag of container to run	latest
-g	Number of GPUs to request	0
-h	Print help	N/A
-i	Name of image to run	N/A
-p	Partition to run image	N/A
-r	Registry to pull the container image	None
-w	Container Work directory where application is	None