- per default running ./prepare.sh will build the model-server image locally
- it takes about 15 minutes to prepare and about 20 minutes to compile tensorflow and the server on a workstation (about 2,5 hours on a vm)
- per default the image will try to allocate 4GB Memory and all CPU there is to bazel during the compilation
- bazel sometimes fails during dependency downloads if a mirror is unreachable, you might just try to re-run
./prepare.sh
after a few minutes