tensorflow serving GPU编译问题

最新推荐文章于 2022-11-10 18:18:29 发布

weixin_34125592

最新推荐文章于 2022-11-10 18:18:29 发布

阅读量546

点赞数

文章标签： python

本文介绍了解决在使用Bazel构建TensorFlow Serving GPU版本过程中遇到的nosuchtarget错误的方法。通过配置环境变量及修改构建配置文件，成功解决了目标未声明的问题。

摘要生成于 C知道，由 DeepSeek-R1 满血版支持，前往体验 >

编译gpu版本：bazel build -c opt --config=cuda --spawn_strategy=standalone //tensorflow_serving/model_servers:tensorflow_model_server

编译cpu版本：bazel build //tensorflow_serving/model_servers:tensorflow_model_server

问题1：

ERROR: no such target '@org_tensorflow//third_party/gpus/crosstool:crosstool': target 'crosstool' not declared in package 'third_party/gpus/crosstool' defined by /home/username/.cache/bazel/_bazel_jorge/2fd988219920b10e9ede8d3b5720f3d2/external/org_tensorflow/third_party/gpus/crosstool/BUILD.

解决：

 1 export TF_NEED_CUDA=1
 2 export TF_NEED_GCP=1
 3 export TF_NEED_JEMALLOC=1
 4 export TF_NEED_HDFS=0
 5 export TF_NEED_OPENCL=0
 6 export TF_ENABLE_XLA=0
 7 export TF_CUDA_VERSION=8.0
 8 export TF_CUDNN_VERSION=5
 9 export TF_CUDA_COMPUTE_CAPABILITIES="3.5,5.2,6.1"
10 export CUDA_TOOLKIT_PATH="/usr/local/cuda"
11 export CUDNN_INSTALL_PATH="/usr/local/cuda"
12 export GCC_HOST_COMPILER_PATH="/usr/bin/gcc"
13 export PYTHON_BIN_PATH="/home/opt/anaconda/envs/py2/bin/python"
14 export CC_OPT_FLAGS="-march=native"
15 export PYTHON_LIB_PATH="/home/opt/anaconda/envs/py2/lib/python2.7/site-packages"
16 
17 cd tensorflow
18 ./configure
19 cd ..
20 
21 # Ref: https://github.com/tensorflow/serving/issues/318#issuecomment-283498443
22 sed -i.bak 's/@org_tensorflow\/\/third_party\/gpus\/crosstool/@local_config_cuda\/\/crosstool:toolchain/g' tools/bazel.rc
23 
24 bazel build -c opt --config=cuda --spawn_strategy=standalone //tensorflow_serving/model_servers:tensorflow_model_server