Python3 的 TF 编译+安装：无法导入名称“resnet”答案

【问题标题】：TF compile+install for Python3: cannot import name 'resnet'Python3 的 TF 编译+安装：无法导入名称“resnet”
【发布时间】：2019-07-30 16:11:31
【问题描述】：

我正在尝试从源代码编译 Tensorflow，因为自 2018 年以来预编译包的 AVX 问题：https://github.com/tensorflow/tensorflow/issues/19584

我在 Fitlet2 MiniPC 上运行 Linux Mint。 “uname -a” 说：

Linux john-fitlet2 4.15.0-55-generic #60-Ubuntu SMP Tue Jul 2 18:22:20 UTC 2019 x86_64 x86_64 x86_64 GNU/Linux

“cat /proc/cpuinfo”说：Intel(R) Celeron(R) CPU J3455 @ 1.50GHz

编译+安装后，当我尝试在 Python3 中导入 tensorflow 时失败：

 ImportError: cannot import name 'resnet'

我做了什么：跟随https://www.tensorflow.org/install/source，但将 pip 更改为 pip3，我执行了如下所示的命令。 “bazel build”行花了大约 24 小时，似乎成功了。

sudo apt install python3-dev python3-pip
pip3 install -U pip six numpy wheel setuptools mock future>=0.17.1
pip3 install -U keras_applications==1.0.6 --no-deps
pip3 install -U keras_preprocessing==1.0.5 --no-deps

git clone https://github.com/tensorflow/tensorflow.git
cd tensorflow
./configure
bazel build --local_ram_resources=2048 --config=opt //tensorflow/tools/pip_package:build_pip_package
./bazel-bin/tensorflow/tools/pip_package/build_pip_package /tmp/tensorflow_pkg
sudo -H pip3 install /tmp/tensorflow_pkg/tensorflow-1.14.0-cp36-cp36m-linux_x86_64.whl

看到问题后，基于在这里用“无法导入名称 resnet”搜索其他问题，我做到了：

sudo -H pip3 install --upgrade keras keras-applications

但我仍然遇到同样的错误，如下所示。

Python 3.6.8 (default, Jan 14 2019, 11:02:34) 
[GCC 8.0.1 20180414 (experimental) [trunk revision 259383]] on linux
Type "help", "copyright", "credits" or "license" for more information.
>>> import tensorflow as tf
Traceback (most recent call last):
  File "<stdin>", line 1, in <module>
  File "/usr/local/lib/python3.6/dist-packages/tensorflow/__init__.py", line 99, in <module>
    from tensorflow_core import *
  File "/usr/local/lib/python3.6/dist-packages/tensorflow_core/__init__.py", line 28, in <module>
    from tensorflow.python import pywrap_tensorflow  # pylint: disable=unused-import
  File "/usr/local/lib/python3.6/dist-packages/tensorflow/__init__.py", line 50, in __getattr__
    module = self._load()
  File "/usr/local/lib/python3.6/dist-packages/tensorflow/__init__.py", line 44, in _load
    module = _importlib.import_module(self.__name__)
  File "/usr/lib/python3.6/importlib/__init__.py", line 126, in import_module
    return _bootstrap._gcd_import(name[level:], package, level)
  File "/usr/local/lib/python3.6/dist-packages/tensorflow_core/python/__init__.py", line 83, in <module>
    from tensorflow.python import keras
  File "/usr/local/lib/python3.6/dist-packages/tensorflow_core/python/keras/__init__.py", line 26, in <module>
    from tensorflow.python.keras import activations
  File "/usr/local/lib/python3.6/dist-packages/tensorflow_core/python/keras/__init__.py", line 27, in <module>
    from tensorflow.python.keras import applications
  File "/usr/local/lib/python3.6/dist-packages/tensorflow_core/python/keras/applications/__init__.py", line 64, in <module>
    from tensorflow.python.keras.applications.resnet import ResNet50
  File "/usr/local/lib/python3.6/dist-packages/tensorflow_core/python/keras/applications/resnet.py", line 22, in <module>
    from keras_applications import resnet
ImportError: cannot import name 'resnet'

【问题讨论】：

标签： python-3.x tensorflow keras compilation installation

【解决方案1】：

解决了。还有一些未知的系统冲突，可能是由于之前安装了 TF 的二进制文件。我卸载了源代码构建的 TF，在https://docs.python-guide.org/dev/virtualenvs/ 之后使用“pipenv”创建了一个虚拟环境并运行：

pip install /tmp/tensorflow_pkg/tensorflow-1.14.0-cp36-cp36m-linux_x86_64.whl

就是这样。如果您忽略来自 numpy 的 FutureWarning 噪音，一切正常：

$ python -c 'import warnings; warnings.simplefilter(action="ignore", category=FutureWarning);import tensorflow as tf; print(tf.__version__)'
1.14.0

【讨论】：