A working installation of Docker Engine with Docker Compose.
To use this neat little bundle of love.. (run commands from the project directory.)
'docker-compose up'
'docker-compose stop
'docker-compose down
All commands need to executed from the root directory of this project.
After starting with docker compose then Then browse to http://localhost:8888
The password is listed in the README.md file in the .jupyter folder - YOU SHOULD CHANGE THIS.
Persistent data stores setup for exiting and new files that can be modified by the container.
I took a copy of the main pytorch image with 'docker pull pytorch/pytorch'
Jumped onto the running containing and installed jupyter from the container commandline like with: conda install -c anaconda jupyter
Then saved changes to the container and pushed it up to my buddhafinger repo.
https://hub.docker.com/r/buddhafinger/pytorch-jupyter/
Once I created and modified the jupyter config file I then wrapped it with docker-compose.
import sys !conda install --yes --prefix {sys.prefix} numpy
#Note - The modules will exist as long as the container does, if you delete the container (not the image) you lose the updated modules, to get around this you can write docker changes to your local image. You can google how to that.
https://www.youtube.com/watch?v=HW29067qVWk
https://www.youtube.com/watch?v=nbJ-2G2GXL0&vl=en
https://pytorch.org/tutorials/beginner/chatbot_tutorial.html
https://pytorch.org/tutorials/beginner/deep_learning_60min_blitz.html