How to create a container from an image

Question

I have this following image:

REPOSITORY               TAG         IMAGE ID       CREATED       SIZE
nvcr.io/nvidia/pytorch   18.04-py3   c3555fec4fe6   3 years ago   5.9GB

The goal is to use Ubuntu 18.04 with CUDA 10.2. Now I want to create a container with this image. How can I do that? The guides online didn't make much sense to me as I'm pretty new to docker.

Answer 1

Depending on what you want to do, there are a lot of options you can pass in to docker, but if you just want to run the code in the container as written:

docker run nvcr.io/nvidia/pytorch:18.04-py3

or

docker run c3555fec4fe6

or even

docker run c3

you can give your docker images a more reasonable name with

docker image tag nvcr.io/nvidia/pytorch:18.04-py3 pytorch

and run that with

docker run pytorch

---

see

• man docker-run

• man docker-image-tag

and https://docs.docker.com/engine/reference/commandline/run/

Answer 2

This would be the multi-layer approach in docker containers...

#Create the base OS image
FROM python:3 AS base

#Update the OS ubuntu image
RUN apt-get -y update

#Install packages
RUN apt-get -y install firefox \
&& apt-get -y install vim

#Create another image layer on top of base to install requirements
FROM base AS requirements

#Install the requirements
RUN pip3 install -r requirements.txt

#Create an intermediate image layer for testing purpose
FROM requirements as test

#Create the build context
COPY /usr/src/my-app /desktop/my-app

#Test the final app
CMD ["python3", "index.py"]

Take a look at this: https://www.geeksforgeeks.org/how-to-combine-multiple-base-images-using-single-dockerfile/