Tensorflow serving提供了部署tensorflow生成的模型给线上服务的方法，包括模型的export，load等等。
安装参考这个
但是由于被qiang的问题 (googlesource无法访问)
需要修改一下 WORKSPACE文件
new_http_archive(
name = "gmock_archive",
#url = "",
url = "",
#sha256 = "26fcbb5925b74ad5fc8c26b0495dfc96353f4d553492eb97e85a8a6d2f43095b",
sha256 = "9b0018413f4222b8ee5454a431918c324d010062eecb078677b6897d0c76bc42",
build_file = "tensorflow/google/protobuf/gmock.BUILD",
)
git_repository(
name = "boringssl_git",
#commit = "436432d849b83ab90f18773e4ae1c7a8f148f48d",
commit = "db0729054d5964feab9e60089ba2d06a181e78b1",
init_submodules = True,
#remote = "",
remote = "",
)
最近发现一个付费vpn，但是更方便一些，免得修改了。。
http://www.lvdou888.com/intro/ 绿豆vpn 挺好用 我在win下开启 可以再ubuntu下面命令行执行bazel 什么访问畅通了。。 还是很赞
注意需要先进入tensorflow路径
./configure一下
bazel build tensorflow_serving/…
最后验证一下
./bazel-bin/tensorflow_serving/example/mnist_inference --port=9000 ~/tmp/
I tensorflow_serving/session_bundle/session_bundle.cc:109] Attempting to load a SessionBundle from: /home/gezi/tmp/
E tensorflow_serving/example/mnist_inference.cc:190] Fail to load tensorflow export: /home/gezi/tmp/export.meta
gezi:~/other/serving$ bazel test tensorflow_serving/...
WARNING: Sandboxed execution is not supported on your system and thus hermeticity of actions cannot be guaranteed. See for more information. You can turn off this warning via --ignore_unsupported_sandboxing.
INFO: Found 89 targets and 36 test targets...
INFO: Elapsed time: 31.664s, Critical Path: 28.86s
//tensorflow_serving/batching:batch_scheduler_test PASSED in 0.3s
//tensorflow_serving/batching:retrier_test PASSED in 0.2s
//tensorflow_serving/batching:streaming_batch_scheduler_test PASSED in 4.1s
//tensorflow_serving/batching/test_util:puppet_batch_scheduler_test PASSED in 0.2s
//tensorflow_serving/core:availability_helpers_test PASSED in 0.3s
//tensorflow_serving/core:dynamic_manager_benchmark PASSED in 28.5s
//tensorflow_serving/core:dynamic_manager_test PASSED in 0.3s
//tensorflow_serving/core:eager_load_policy_test PASSED in 0.2s