Bazel项目+Tensorflow依赖关系问题
- 首先，在WORKSPACE文件中添加TensorFlow依赖项
load("@bazel_tools//tools/build_defs/repo:http.bzl", "http_archive")
http_archive( name = "io_bazel_rules_docker", urls = ["https://github.com/bazelbuild/rules_docker/releases/download/v0.6.1/rules_docker-v0.6.1.tar.gz"], sha256 = "a6b1ebb29bae44e2aa8ca2db7939d62aa07d585f7f4008f0353d6eb6d62b9bf7", )
http_archive( name = "org_tensorflow", urls = ["https://mirror.bazel.build/github.com/tensorflow/tensorflow/archive/v2.4.1.tar.gz"], sha256 = "d9285d6f2b673ece17c86b01132f84b7bea9c757a4f91706a883ce33897d4a9b", strip_prefix = "tensorflow-2.4.1", )
- 接下来，确保运行Bazel构建的Python可执行文件可以找到TensorFlow
python_binary( name = "my_executable", main = "my_executable.py", srcs = ["my_executable.py"], deps = [ "//tensorflow:tensorflow", "@numpy//:numpy", ], )
- 最后，确保在运行Bazel构建之前安装TensorFlow
$ pip install tensorflow==2.4.1
或者使用Bazel提供的TensorFlow编译选项
$ bazel build --config=opt //tensorflow/tools/pip_package:build_pip_package $ ./bazel-bin/tensorflow/tools/pip_package/build_pip_package /tmp/tensorflow_pkg
然后再使用pip来安装
$ pip install /tmp/tensorflow_pkg/tensorflow-2.4.1-*.whl