Question

尝试从源代码编译TensorFlow时出现以下错误。任何想法都会有所帮助。

bazel-out/host/bin/_solib_local/_U_S_Stensorflow_Spython_Cgen_Unn_Uops_Upy_Uwrappers_Ucc___Utensorflow/libtensorflow_framework.so: undefined reference to `cublasGemmEx@libcublas.so.9.0'
bazel-out/host/bin/_solib_local/_U_S_Stensorflow_Spython_Cgen_Unn_Uops_Upy_Uwrappers_Ucc___Utensorflow/libtensorflow_framework.so: undefined reference to `cublasZhpmv_v2@libcublas.so.9.0'
bazel-out/host/bin/_solib_local/_U_S_Stensorflow_Spython_Cgen_Unn_Uops_Upy_Uwrappers_Ucc___Utensorflow/libtensorflow_framework.so: undefined reference to `cufftExecD2Z@libcufft.so.9.0'
bazel-out/host/bin/_solib_local/_U_S_Stensorflow_Spython_Cgen_Unn_Uops_Upy_Uwrappers_Ucc___Utensorflow/libtensorflow_framework.so: undefined reference to `cublasSrotg_v2@libcublas.so.9.0'
bazel-out/host/bin/_solib_local/_U_S_Stensorflow_Spython_Cgen_Unn_Uops_Upy_Uwrappers_Ucc___Utensorflow/libtensorflow_framework.so: undefined reference to `cufftExecR2C@libcufft.so.9.0'
bazel-out/host/bin/_solib_local/_U_S_Stensorflow_Spython_Cgen_Unn_Uops_Upy_Uwrappers_Ucc___Utensorflow/libtensorflow_framework.so: undefined reference to `cublasSsyrk_v2@libcublas.so.9.0'
bazel-out/host/bin/_solib_local/_U_S_Stensorflow_Spython_Cgen_Unn_Uops_Upy_Uwrappers_Ucc___Utensorflow/libtensorflow_framework.so: undefined reference to `cublasDgemm_v2@libcublas.so.9.0'
bazel-out/host/bin/_solib_local/_U_S_Stensorflow_Spython_Cgen_Unn_Uops_Upy_Uwrappers_Ucc___Utensorflow/libtensorflow_framework.so: undefined reference to `cufftSetWorkArea@libcufft.so.9.0'
bazel-out/host/bin/_solib_local/_U_S_Stensorflow_Spython_Cgen_Unn_Uops_Upy_Uwrappers_Ucc___Utensorflow/libtensorflow_framework.so: undefined reference to `cublasChemm_v2@libcublas.so.9.0'
bazel-out/host/bin/_solib_local/_U_S_Stensorflow_Spython_Cgen_Unn_Uops_Upy_Uwrappers_Ucc___Utensorflow/libtensorflow_framework.so: undefined reference to `cublasZher2k_v2@libcublas.so.9.0'
bazel-out/host/bin/_solib_local/_U_S_Stensorflow_Spython_Cgen_Unn_Uops_Upy_Uwrappers_Ucc___Utensorflow/libtensorflow_framework.so: undefined reference to `cufftExecC2C@libcufft.so.9.0'
bazel-out/host/bin/_solib_local/_U_S_Stensorflow_Spython_Cgen_Unn_Uops_Upy_Uwrappers_Ucc___Utensorflow/libtensorflow_framework.so: undefined reference to `curandSetStream@libcurand.so.9.0'
bazel-out/host/bin/_solib_local/_U_S_Stensorflow_Spython_Cgen_Unn_Uops_Upy_Uwrappers_Ucc___Utensorflow/libtensorflow_framework.so: undefined reference to `cublasDrotm_v2@libcublas.so.9.0'
bazel-out/host/bin/_solib_local/_U_S_Stensorflow_Spython_Cgen_Unn_Uops_Upy_Uwrappers_Ucc___Utensorflow/libtensorflow_framework.so: undefined reference to `curandSetPseudoRandomGeneratorSeed@libcurand.so.9.0'

Answer 1

我们的构建中似乎存在错误。我能够在我的机器上重现相同的内容。看起来LD_LIBRARY_PATH的值并不总是在bazel构建期间正确传播。就我而言，当我使用此命令时，我能够成功构建：

bazel build  --config=opt --config=cuda tensorflow/tools/pip_package:build_pip_package --action_env="LD_LIBRARY_PATH=${LD_LIBRARY_PATH}"

Answer 2

我昨天遇到同样的错误，同时试图从源头建立一个显然有效的cuda 9.0的张量流。就我而言，git clean和action_env的任何组合都没有帮助 - ld通过bazel会一直拒绝承认cuda libs。

我最终按照this thread中的说明操作：以root身份创建一个文件/etc/ld.so.conf.d/cuda.conf，其中包含一行

/usr/local/cuda/lib64

（假设您的/usr/local/cuda/与您的具体cuda目录相关联，例如/usr/local/cuda-9.0/。）

然后发出sudo ldconfig。有了这个，构建就完成了，而tensorflow正在使用我的GPU。

Answer 3

试图让这个问题更容易搜索：我收到的错误信息也包含在顶部：

libcublas.so.9.0, needed by bazel-out/[...]/libtensorflow_framework.so, not found (try using -rpath or -rpath-link)

对于libcudnn等

等等。

当我遇到此问题时，我首先将/usr/local/cuda/lib64和/usr/local/cuda/extras/CUPTI/lib64添加到LD_LIBRARY_PATH并尝试重建（不使用--action_env）。没有工作。

然后我再次进行了一次干净的重新配置和构建，没有--action_env，并且它有效。我通过git clean -xdf清理了我的存储库，谨慎地说，这将会破坏存储库中git不知道的所有文件。：）

也许--action_env会避免进行干净的重建，我不知道。但是，如果在进行第一次构建之前，这些库位于LD_LIBRARY_PATH，我希望您不需要--action_env。

Answer 4

在发生错误之后，我在/usr/local/cuda/lib64后面附加了LD_LIBRARY_PATH。没用然后，我再次用.tf_configure.bazelrc修改了build --action_env LD_LIBRARY_PATH=..."。重新编译项目并通过！

TensorFlow无法编译

4 个答案: