Tensorflow：ImportError无法在retrain.py中导入名称graph_util

Question

我从源代码编译了Tensorflow以使用GPU。我能够成功安装它。它能够成功运行示例mnist convolutional.py脚本。

但是，当我在中执行retrain.py脚本时tensorflow/examples/image_retraining，它会吐出以下错误：

Traceback (most recent call last):
  File "retrain.py", line 70, in <module>
    from tensorflow.python.client import graph_util
ImportError: cannot import name graph_util

带有命令和其他输出：

$ python retrain.py \
--bottleneck_dir=gpu_train/bottlenecks \
--model_dir=inception \
--output_graph=gpu_train/retrained_graph.pb \
--output_labels=gpu_train/retrained_labels.txt \
--image_dir categories
I tensorflow/stream_executor/dso_loader.cc:108] successfully opened CUDA library libcublas.dylib locally
I tensorflow/stream_executor/dso_loader.cc:108] successfully opened CUDA library libcudnn.dylib locally
I tensorflow/stream_executor/dso_loader.cc:108] successfully opened CUDA library libcufft.dylib locally
I tensorflow/stream_executor/dso_loader.cc:108] successfully opened CUDA library libcuda.dylib locally
I tensorflow/stream_executor/dso_loader.cc:108] successfully opened CUDA library libcurand.dylib locally
Traceback (most recent call last):
  File "retrain.py", line 70, in <module>
    from tensorflow.python.client import graph_util
ImportError: cannot import name graph_util

Answer 1

昨天有一次提交，将文件graph_util.py从tensorflow.python.client移到tensorflow.python.framework。

您retrain.py使用import 的旧版本很奇怪from tensorflow.python.client import graph_util，但是也graph_util.py搬了。

无论如何，主版本的graph_util.py是最新的（有from tensorflow.python.framework import graph_util），所以你应该得到的最新版本（无需重新编译，只是拉）。