Create a dockerfile for coastseg

[2320sharon]

CoastSeg relies on tensorflow to run the zoo workflow, but not the coastsat workflow. This makes tensorflow and the transformers library optional dependencies. However if users on Windows want to install tensorflow it can be an absolute nightmare since tensorflow hasn't been fully compatible with windows since tensorflow gpu.

This issue has been created to track all development on the creation of a docker file that can run coastseg and the rest of the doodleverse.

This dockerfile needs to be able to:

• access the GPU
• install conda packages
• pip packages

Files

We were able to make some initial progress on this with the help of Venus. Here are the starter dockerfiles and scripts
Dockerfile_og.zip : originally created by @venuswku
adjust_permissions.zip
Dockerfile.zip

Commands

The following commands I ran in the command prompt on windows to get it running
The command I used to build the docker image from the docker file docker build -t coastseg_slim_3 .
Note: When I found this dockerfile it had already been built a few months back so the version of coastseg is pretty old
The command I used to run the docker image docker run -p 8888:8888 coastseg_slim_3

Resources

Here are the resources I had saved while I was working on this
Jupyter Official notebook docker images
This looks like the solution
https://jupyter-docker-stacks.readthedocs.io/en/latest/using/recipes.html#:~:text=as root.-,Using mamba install (recommended) or pip install in a Child Docker image,-%23
Base docker file
https://github.com/jupyter/docker-stacks/blob/a5b40a6f1117bd675565b3673e063125dd74eac3/images/base-notebook/Dockerfile
Geemap’s dockerfile
https://github.com/gee-community/geemap/blob/master/Dockerfile
Zip