Как скомпилировать TF-Обслуживание с локальным TF?
TF commit - 184388ce4c5290b26b11ffcf860fe633772a0598, TF 2.1
TF-Serving commit 936542b1041e4ed063ce747cc1b0d5f65d15fb02, Msgstr "Обновить версию для выпуска 2.1.0-rc1."
Я уже читал эту базу данных тензорного потока, использующую локальный загруженный тензор потока . Но этот пост не может решить мою проблему.
Что я делаю: В WORKSPACE я меняю tensorflow_http_archive на local_repository. Однако я не могу скомпилировать без или без run_in_docker.sh.
1 workspace(name = "tf_serving")
2
3 # To update TensorFlow to a new revision.
4 # 1. Update the 'git_commit' args below to include the new git hash.
5 # 2. Get the sha256 hash of the archive with a command such as...
6 # curl -L https://github.com/tensorflow/tensorflow/archive/<git hash>.tar.gz | sha256sum
7 # and update the 'sha256' arg with the result.
8 # 3. Request the new archive to be mirrored on mirror.bazel.build for more
9 # reliable downloads.
10 #load("//tensorflow_serving:repo.bzl", "tensorflow_http_archive")
11 #tensorflow_http_archive(
12 # name = "org_tensorflow",
13 # sha256 = "6cb1da6b285db2d486510c20896c6e6dc0ac70476af58bfb0e70b82643dc0406",
14 # git_commit = "064e1535a7ba3cb9f67b9d9171309a1e9ebca2b0",
15 #)
16 local_repository(
17 name="org_tensorflow",
18 path="/home/wxf/tensorflow",
19 )
Без run_in_docker.sh:
:~/tf2/serving$ bazel build -c opt tensorflow_serving/...
ERROR: /home/wxf/tf2/serving/tensorflow_serving/apis/BUILD:103:1: in deps attribute of cc_library rule //tensorflow_serving/apis:predict_proto: proto_library rule '@org_tensorflow//tensorflow/core:protos_all' is misplaced here (expected cc_library, objc_library, cc_proto_library or cc_import) and '@org_tensorflow//tensorflow/core:protos_all' does not have mandatory providers: 'CcInfo'. Since this rule was created by the macro 'serving_proto_library', the error might have been caused by the macro implementation
ERROR: Analysis of target '//tensorflow_serving/apis:predict_proto' failed; build aborted: Analysis of target '//tensorflow_serving/apis:predict_proto' failed; build aborted
INFO: Elapsed time: 0.372s
INFO: 0 processes.
FAILED: Build did NOT complete successfully (0 packages loaded, 0 targets configured)
Я могу предоставить больше информации, если это необходимо. Помогите пожалуйста дать пошаговый способ сборки из sr c. Спасибо.
Ссылка
[1] https://www.tensorflow.org/tfx/serving/setup#building_from_source
[2] https://github.com/tensorflow/serving