build TensorFlow error

daxinnan · 2018 年10 月 23 日 05:16


julia> Pkg.status()
    Status `~/.julia/environments/v1.0/Project.toml`
  [8f4d0f93] Conda v1.1.1
  [438e738f] PyCall v1.18.5
  [1d978283] TensorFlow v0.10.2

julia> versioninfo()
Julia Version 1.0.1
Commit 0d713926f8* (2018-09-29 19:05 UTC)
Platform Info:
  OS: Linux (x86_64-redhat-linux)
  CPU: Intel(R) Xeon(R) CPU E5-2680 v4 @ 2.40GHz
  WORD_SIZE: 64
  LIBM: libopenlibm
  LLVM: libLLVM-6.0.0 (ORCJIT, broadwell)

我在build时候遇到一个网络无法访问的问题，如何能解决呢？因为在大型机上跑，也并不能更改hosts啥的。

然后我在自己电脑上下好把压缩包上传到那个路径之后并不起作用，他还是会下载。然后出现问题。
我在github的TensorFlow.jl提交了issue

github.com/malmaud/TensorFlow.jl

build TensorFlow Error

opened 03:21AM - 22 Oct 18 UTC

closed 08:59AM - 26 Oct 18 UTC

daxinnan

``` julia> ENV["PYTHON"]="" "" julia> using Pkg julia> Pkg.build("TensorFl…ow") Building Arpack ──────────→ `~/.julia/packages/Arpack/UiiMc/deps/build.log` Building SpecialFunctions → `~/.julia/packages/SpecialFunctions/sXbz6/deps/build.log` Building Rmath ───────────→ `~/.julia/packages/Rmath/Py9gH/deps/build.log` Building Conda ───────────→ `~/.julia/packages/Conda/uQitS/deps/build.log` Building MbedTLS ─────────→ `~/.julia/packages/MbedTLS/mkHpa/deps/build.log` Building PyCall ──────────→ `~/.julia/packages/PyCall/0jMpb/deps/build.log` Building CodecZlib ───────→ `~/.julia/packages/CodecZlib/wwgbh/deps/build.log` Building TensorFlow ──────→ `~/.julia/packages/TensorFlow/A6TdG/deps/build.log` ┌ Error: Error building `TensorFlow`: │ Warning: 'conda-forge' already in 'channels' list, moving to the top │ Solving environment: ...working... done │ │ # All requested packages already installed. │ │ % Total % Received % Xferd Average Speed Time Time Time Current │ Dload Upload Total Spent Left Speed 0 0 0 0 0 0 0 0 --:--:-- 0:06:01 --:--:-- 0 │ curl: (56) TCP connection reset by peer │ [ Info: Building TensorFlow.jl for CPU use only. To enable the GPU, set the TF_USE_GPU environment variable to 1 and rebuild TensorFlow.jl │ [ Info: Running `conda config --add channels conda-forge --force` in root environment │ [ Info: Running `conda install -y tensorflow=1.10.0` in root environment │ ERROR: LoadError: failed process: Process(`curl -g -L -f -o /xxxxx/.julia/packages/TensorFlow/A6TdG/deps/downloads/tensorflow.tar.gz https://storage.googleapis.com/tensorflow/libtensorflow/libtensorflow-cpu-linux-x86_64-1.10.0.tar.gz`, ProcessExited(56)) [56] │ Stacktrace: │ [1] error(::String, ::Base.Process, ::String, ::Int64, ::String) at ./error.jl:42 │ [2] pipeline_error at ./process.jl:705 [inlined] │ [3] #run#504(::Bool, ::Function, ::Cmd) at ./process.jl:663 │ [4] run at ./process.jl:661 [inlined] │ [5] download(::String, ::String) at ./download.jl:27 │ [6] download_and_unpack(::String) at /xxxxxx/.julia/packages/TensorFlow/A6TdG/deps/build.jl:78 │ [7] top-level scope at /xxxxxx/.julia/packages/TensorFlow/A6TdG/deps/build.jl:99 │ [8] include at ./boot.jl:317 [inlined] │ [9] include_relative(::Module, ::String) at ./loading.jl:1041 │ [10] include(::Module, ::String) at ./sysimg.jl:29 │ [11] include(::String) at ./client.jl:388 │ [12] top-level scope at none:0 │ in expression starting at /xxxxxx/.julia/packages/TensorFlow/A6TdG/deps/build.jl:93 └ @ Pkg.Operations /xxxx/download/julia/usr/share/julia/stdlib/v1.0/Pkg/src/Operations.jl:1069 ``` The network is not so good,and I don't have enough permission to change anything. What can I do for this, is there any offline way to solve it? I download `tensorflow.tar.gz` by `curl -g -L -f -o /xx/tensorflow.tar.gz https://storage.googleapis.com/tensorflow/libtensorflow/libtensorflow-cpu-linux-x86_64-1.10.0.tar.gz` in another network environment and move the tarball to the required location, but it doesn't work. my versioninfo is ``` julia> versioninfo() Julia Version 1.0.1 Commit 0d713926f8* (2018-09-29 19:05 UTC) Platform Info: OS: Linux (x86_64-redhat-linux) CPU: Intel(R) Xeon(R) CPU E5-2680 v4 @ 2.40GHz WORD_SIZE: 64 LIBM: libopenlibm LLVM: libLLVM-6.0.0 (ORCJIT, broadwell) ```

但是昨天github有问题，被我不小心都close了。今天重新开的，有小伙伴遇到过这个问题吗，有没有合适的解决办法。
刚刚学校老师说，学校大型机没有配ipv6

Roger · 2018 年10 月 23 日 07:22

你被墙了，tf本身是被墙的，conda默认会从Google的官方repo下载。所以你可能需要自己从某个境内镜像手动下载 tar.gz 然后复制到 curl 的那个目录里面去

daxinnan · 2018 年10 月 23 日 09:37

关键是现在放进去不起作用，它还会执行这一句。。。

Rain · 2018 年10 月 27 日 02:29

fq
在 hosts 将 storage.googleapis.com 设为 127.0.0.1 ，然后使用 nginx。
找到源码进行修改
看报错信息，应该是使用 conda 安装 tensorflow 时报错，能不能先在终端中调用 conda 安装，然后再使用 julia 进行 build 操作。（这个好像根本就不可行）

Roger · 2018 年10 月 27 日 17:47

作者已经在issue里面给了解决方法了：

https://github.com/malmaud/TensorFlow.jl/issues/453

@daxinnan mark一下已解决吧，点个勾就OK

daxinnan · 2018 年10 月 29 日 02:42

是因为时间超了吗，这个帖子并不像我发的其他帖子可以看到已解决的框选，我无法设置已解决。

Gnimuc · 2018 年10 月 30 日 01:39

之前这个区没开启标记解决，现在开了。