I'm trying to automate the build process of a tensorflow serving docker image, adding the bazel-build for the model server in the provided Dockerfile.
The problem is after cloning the serving repo we have to configure it, and while most configuration steps can be pre-answered with env variables like
RUN export PYTHON_BIN_PATH=/usr/bin/python && \
export TF_NEED_GCP=0 && \
export TF_NEED_HDFS=0
The configuration of the python libs controlled by util/python/python_config.sh can only be achieved with normal input.
I know this is actually an issue in tensorflow instead of serving, if I make a PR there I'll reference this one.
I'm trying to automate the build process of a tensorflow serving docker image, adding the
bazel-buildfor the model server in the provided Dockerfile.The problem is after cloning the
servingrepo we have to configure it, and while most configuration steps can be pre-answered with env variables likeThe configuration of the python libs controlled by
util/python/python_config.shcan only be achieved with normal input.I know this is actually an issue in
tensorflowinstead ofserving, if I make a PR there I'll reference this one.