The Dockerfiles in this directory build images containing the
minimum necessary to run, build, or interact with the NVVL library
through pytorch. They probably won't be very useful as-is but serve
as a base for expanding or examples to take from. While these are
similar to the Dockerfile's in the base Docker directory, the base
pytorch images have more packages already installed than the base cuda
images, so we need to install less. Also, we don't include _opencv
options since it is assumed pytorch will be used for anything opencv
would have been.
-
Dockerfile.minimal- The minimum necessary to use the nvvl pytorch library. -
Dockerfile.build- Includes tools necessary to build the library. -
Dockerfile.interactive- Includes tools we found useful for interactive use (mostly theffmpegcommandline tool)
The final two are useful for the decidedly non-docker-ish practice of developing while inside a container.