本文主要是介绍tensorflow 1.15编译，希望对大家解决编程问题提供一定的参考价值，需要的开发者们随着小编来一起学习吧！
update，最新的1.15.5版本死链接已经修复了。
----------------------------------------
tensorflow真的是烂，build from source一堆死链接死依赖，官方连自己release版本的availability都不管，难怪弃坑的人越来越多
先改这几个文件
thw@thw-s5:~/code/tf-gitee/tensorflow$ git diff
diff --git a/tensorflow/workspace.bzl b/tensorflow/workspace.bzl
index 84b2bdcb2d..615d8c3cc3 100755
--- a/tensorflow/workspace.bzl
+++ b/tensorflow/workspace.bzl
@@ -135,11 +135,14 @@ def tf_repositories(path_prefix = "", tf_repo_name = ""):tf_http_archive(name = "mkl_dnn",build_file = clean_dep("//third_party/mkl_dnn:mkldnn.BUILD"),
- sha256 = "74675e93eef339ff3d9a9be95c15d0c7ad8736a5356c23428ab2e33dcdb8e3e1",
- strip_prefix = "mkl-dnn-0.20.6",
+ #sha256 = "74675e93eef339ff3d9a9be95c15d0c7ad8736a5356c23428ab2e33dcdb8e3e1",
+ sha256 = "f1de676fddeb94132c5ae480fb03a64f03deda10b09b3141373f2b2fe5cd031d",
+ #strip_prefix = "mkl-dnn-0.20.6",
+ strip_prefix = "oneDNN-0.20.6",urls = ["https://storage.googleapis.com/mirror.tensorflow.org/github.com/intel/mkl-dnn/archive/v0.20.6.tar.gz","https://github.com/intel/mkl-dnn/archive/v0.20.6.tar.gz",
+ "https://github.com/oneapi-src/oneDNN/archive/v0.20.6.tar.gz",],)@@ -409,9 +412,11 @@ def tf_repositories(path_prefix = "", tf_repo_name = ""):name = "org_python_license",licenses = ["notice"], # Python 2.0sha256_urls = {
- "e76cacdf0bdd265ff074ccca03671c33126f597f39d0ed97bc3e5673d9170cf6": [
- "https://storage.googleapis.com/mirror.tensorflow.org/docs.python.org/2.7/_sources/license.rst.txt",
- "https://docs.python.org/2.7/_sources/license.rst.txt",
+ #"e76cacdf0bdd265ff074ccca03671c33126f597f39d0ed97bc3e5673d9170cf6": [
+ "b5556e921715ddb9242c076cae3963f483aa47266c5e37ea4c187f77cc79501c": [
+ #"https://storage.googleapis.com/mirror.tensorflow.org/docs.python.org/2.7/_sources/license.rst.txt",
+ #"https://docs.python.org/2.7/_sources/license.rst.txt",
+ "https://mirror.bazel.build/docs.python.org/2.7/_sources/license.txt",],},)
编译脚本
bazel build --config=opt \--config=cuda \--config=nonccl \--config=nokafka \--config=noaws \--config=nogcp \--config=nohdfs \--config=noignite \--jobs 6 \//tensorflow/tools/pip_package:build_pip_package
总算是编译过了
如果不加noxxx那一坨，就会遇到更多死链接
这篇关于tensorflow 1.15编译的文章就介绍到这儿，希望我们推荐的文章对编程师们有所帮助！
