One can build a MLModelScope agent by cd $GOPATH/github.com/rai-project/${framework}/${framework}-agent && go build
and then run ./agent -h
for the usage information. Different usages are available as subcommands to each agent. The command line interface of the agents is built using cobra.
For example for TensorFlow, run
→ cd $GOPATH/src/github.com/rai-project/tensorflow/tensorflow-agent
→ go build
→ ./tensorflow-agent -h
One can run a MLModelScope agent as a wrapper around frameworks to evaluate models locally, or as a model serving process to handle remote evaluation requests from other processes.