Skip to content

Instantly share code, notes, and snippets.

@obinsc

obinsc/build_ts_serving_source.sh

Forked from waichee/build_ts_serving_source.sh
Created May 30, 2017 13:54
Show Gist options

  • Select an option

    Learn more about clone URLs
  • Save obinsc/c34217c90251a18bf0a1088cc3cb84dd to your computer and use it in GitHub Desktop.

Select an option

Learn more about clone URLs
Save obinsc/c34217c90251a18bf0a1088cc3cb84dd to your computer and use it in GitHub Desktop.
Download ZIP
Code to Build Tensorflow Serving from source within a Docker container
Raw
build_ts_serving_source.sh
mkdir -p /work/
# Clone the source from Github
cd /work/ && git clone — recurse-submodules https://github.com/tensorflow/serving
# Pin the version of Tensorflow Serving and its submodule
TENSOR_SERVING_COMMIT_HASH=85db9d3
TENSORFLOW_COMMIT_HASH=dbe5e17
cd /work/serving && git checkout $TENSOR_SERVING_COMMIT_HASH
cd /work/serving/tensorflow && git checkout $TENSORFLOW_COMMIT_HASH
cd /work/serving/tensorflow && ./configure
# Tensorflow Serving uses Bazel as the build tool. The Docker image already have Bazel installed in it.
# Run the following command to build the source with Bazel
cd /work/serving && bazel build -c opt //tensorflow_serving/model_servers:tensorflow_model_server --jobs 10 --curses no --discard_analysis_cache
# The following bazel option flags was added:
# -c (compilation_mode): the compilation mode flag affects the the C++ generation. ‘opt’ compilation mode is selected to enable optimization and disable the assert calls.
# — discard_analysis_cache: will discard the analysis cache immediately after the analysis phase completes. This reduces memory usage by ~10%, but makes further incremental builds slower. 
# — jobs: The default number of jobs spawned by bazel is 200. Depending on the system configuration of your host, you might like to update this parameter. We tune ours to 10.
Sign up for free to join this conversation on GitHub. Already have an account? Sign in to comment