Skip to content

Latest commit

 

History

History
52 lines (34 loc) · 2.48 KB

recommended_workflow.md

File metadata and controls

52 lines (34 loc) · 2.48 KB
Raw

Recommended Workflow

This doc outlines the recommended workflow for pytorch-dense-correspondence.

The typical workflow we use is:

  • edit code which is outside the docker container (so you can use whichever text editor / IDE you like) but that is externally mounted into the docker container
  • to run code, first launch a Jupyter notebook server from inside the docker container
  • run code and visualize data in web browser via Jupyter notebook

One time setup: setting the path to your data directory

Edit config/docker_run_config.yaml to set the path to your data volume. By data volume we mean the folder that contains the pdc data folder as outline in the data organization doc. This will be mounted inside the docker at the location ~/code/data_volume.

Launching a docker container

You should already have built the docker image as outlined here. To launch the container

cd docker
./docker_run.py

Sourcing environment variables

You will need to set a few environment variables to add our modules to the PYTHONPATH. There are two aliased commands, use_pytorch_dense_correspondence and use_director.

  • use_pytorch_dense_correspondence adds several modules to the PYTHONPATH. You will always need to source this.
  • use_director adds director to the path. Make sure run this if using any scripts that rely on director.

Launching a Jupyter notebook server inside the docker container

You are now inside the docker container. Run terminator & you will launch a new terminator with a blue background.

  • source the necessary environment variables with use_pytorch_dense_correspondence. If using scripts that rely on director then also run use_director.
  • start a Jupyter notebook server with the ~/code/start_notebook.py.

Interacting via Jupyter notebook

The terminal from which you ran the start_notebook.py command will print out a URL.

Navigate there to interact via Jupyter.

Docker Cheatsheet

You won't need to know much about docker to work with this repo, since the image can remain untouched once it's built, but here are a few quick commands to know for deleting unused images and containers:

Handling images

  • docker images - lists all docker images on machine, including REPOSITORY, TAG, IMAGE_ID, when created, size
  • docker rmi IMAGE_ID - removes this image

Handling containers

  • docker ps -a - lists all containers on machine
  • docker rm CONTAINER_ID - removes container id