Tensorflow freeze_graph IndexError：列表索引超出范围

Question

我训练了一个用于图像分类的 resnet 模型，并希望冻结图形。当我使用 freeze_graph 工具时，它给了我以下错误：

    Traceback (most recent call last):
  File "/media/goodvirus/34e9f1a6-6a30-4087-b582-b02e43a9aa82/DeepLearning/Entwicklungen/venv/bin/freeze_graph", line 11, in <module>
    sys.exit(run_main())
  File "/media/goodvirus/34e9f1a6-6a30-4087-b582-b02e43a9aa82/DeepLearning/Entwicklungen/venv/local/lib/python2.7/site-packages/tensorflow/python/tools/freeze_graph.py", line 408, in run_main
    app.run(main=my_main, argv=[sys.argv[0]] + unparsed)
  File "/media/goodvirus/34e9f1a6-6a30-4087-b582-b02e43a9aa82/DeepLearning/Entwicklungen/venv/local/lib/python2.7/site-packages/tensorflow/python/platform/app.py", line 125, in run
    _sys.exit(main(argv))
  File "/media/goodvirus/34e9f1a6-6a30-4087-b582-b02e43a9aa82/DeepLearning/Entwicklungen/venv/local/lib/python2.7/site-packages/tensorflow/python/tools/freeze_graph.py", line 407, in <lambda>
    my_main = lambda unused_args: main(unused_args, flags)
  File "/media/goodvirus/34e9f1a6-6a30-4087-b582-b02e43a9aa82/DeepLearning/Entwicklungen/venv/local/lib/python2.7/site-packages/tensorflow/python/tools/freeze_graph.py", line 301, in main
    flags.saved_model_tags, checkpoint_version)
  File "/media/goodvirus/34e9f1a6-6a30-4087-b582-b02e43a9aa82/DeepLearning/Entwicklungen/venv/local/lib/python2.7/site-packages/tensorflow/python/tools/freeze_graph.py", line 283, in freeze_graph
    checkpoint_version=checkpoint_version)
  File "/media/goodvirus/34e9f1a6-6a30-4087-b582-b02e43a9aa82/DeepLearning/Entwicklungen/venv/local/lib/python2.7/site-packages/tensorflow/python/tools/freeze_graph.py", line 145, in freeze_graph_with_def_protos
    var_list=var_list, write_version=checkpoint_version)
  File "/media/goodvirus/34e9f1a6-6a30-4087-b582-b02e43a9aa82/DeepLearning/Entwicklungen/venv/local/lib/python2.7/site-packages/tensorflow/python/training/saver.py", line 1078, in __init__
    self.build()
  File "/media/goodvirus/34e9f1a6-6a30-4087-b582-b02e43a9aa82/DeepLearning/Entwicklungen/venv/local/lib/python2.7/site-packages/tensorflow/python/training/saver.py", line 1090, in build
    self._build(self._filename, build_save=True, build_restore=True)
  File "/media/goodvirus/34e9f1a6-6a30-4087-b582-b02e43a9aa82/DeepLearning/Entwicklungen/venv/local/lib/python2.7/site-packages/tensorflow/python/training/saver.py", line 1127, in _build
    build_save=build_save, build_restore=build_restore)
  File "/media/goodvirus/34e9f1a6-6a30-4087-b582-b02e43a9aa82/DeepLearning/Entwicklungen/venv/local/lib/python2.7/site-packages/tensorflow/python/training/saver.py", line 765, in _build_internal
    saveables = self._ValidateAndSliceInputs(names_to_saveables)
  File "/media/goodvirus/34e9f1a6-6a30-4087-b582-b02e43a9aa82/DeepLearning/Entwicklungen/venv/local/lib/python2.7/site-packages/tensorflow/python/training/saver.py", line 672, in _ValidateAndSliceInputs
    for converted_saveable_object in self.SaveableObjectsForOp(op, name):
  File "/media/goodvirus/34e9f1a6-6a30-4087-b582-b02e43a9aa82/DeepLearning/Entwicklungen/venv/local/lib/python2.7/site-packages/tensorflow/python/training/saver.py", line 646, in SaveableObjectsForOp
    variable, "", name)
  File "/media/goodvirus/34e9f1a6-6a30-4087-b582-b02e43a9aa82/DeepLearning/Entwicklungen/venv/local/lib/python2.7/site-packages/tensorflow/python/training/saver.py", line 128, in __init__
    self.handle_op = var.op.inputs[0]
  File "/media/goodvirus/34e9f1a6-6a30-4087-b582-b02e43a9aa82/DeepLearning/Entwicklungen/venv/local/lib/python2.7/site-packages/tensorflow/python/framework/ops.py", line 2106, in __getitem__
    return self._inputs[i]
IndexError: list index out of range

我使用以下命令启动 freeze_graph：

freeze_graph --input_graph=graph.pbtxt --input_checkpoint=model.ckpt-6300 --output_graph=frozen_graph.pb --output_node_name="softmax"

我在 gpu 支持下运行以下 tensorflow 版本：1.11.0-dev20180808

我还尝试了其他版本的有或没有 gpu 支持，但总是有相同的结果。

云有人指点一下。

此致，

保罗

Answer 1

为了社区的利益，在答案部分提供解决方案。

冻结图像分类 Resnet 模型图的正确命令如下所示：

freeze_graph --input_graph=graph.pbtxt --input_checkpoint=model.ckpt-6300 --output_graph=frozen_graph.pb --output_node_name="softmax_tensor"