I think Docker support might be helpful - especially since many Deep Learning practitioners using this might not be familiar with the Go tool chain.
Imagining an executable image like docker run tarp:latest <tarp-command-here>. Building the image could be part of the future CI. The final image could be as lean as just containing the compiled binary. Base image e.g. go:1.17-alpine.
I think Docker support might be helpful - especially since many Deep Learning practitioners using this might not be familiar with the Go tool chain.
Imagining an executable image like
docker run tarp:latest <tarp-command-here>. Building the image could be part of the future CI. The final image could be as lean as just containing the compiled binary. Base image e.g.go:1.17-alpine.