I’ve added a Dockerfile and submitted it in patch #11
It is paired up with a Makefile.
A few notes:
- The image requires additional capabilities due to swift and lldb.
- The image was not tested with reduced capabilities when using swift-jupyter. Maybe we can sacrifice swift cli repl for better security.
- A makefile was included to help build and start the swiftai environment.
- For now, this image uses the CPU images because GPU on docker requires the installation of docker-nvidia. I’m developing on an OSX laptop which has an AMD GPU, so…
. It would be good to get someone to contribute in this space since there is significant benefit. I personally will probably just skip directly to kubernetes w/GPU support since I have access to that, but that likely won’t help most users getting started.
# Default target builds docker image
make
# Start jupyter with S4TF support
make jupyter
# Start a shell
make shell
You’ll see something like this:
To access the notebook, open this file in a browser:
file:///root/.local/share/jupyter/runtime/nbserver-7-open.html
Or copy and paste one of these URLs:
http://cddd24d4ed29:8888/?token=[TOKEN_ID]
or http://127.0.0.1:8888/?token=[TOKEN_ID]
Copy and paste the 127.0.0.1 (IPv4) version to your browser and have fun!