将 FaceNet model 转换为 ONNX 格式时出错

Question

系统信息

• 操作系统平台和发行版：Linux Ubuntu 19.10
• Tensorflow 版本：1.15
• Python 版本：3.7

问题

我从这个页面下载了一个 tensorflow model 的 FaceNet，我试图将它从 .pb 转换成一个 .onnx 文件，但是它引发了以下错误：

重现

root@xesk-VirtualBox:/home/xesk/Desktop# python -m tf2onnx.convert --saved-model home/xesk/Desktop/2s/20180402-114759/20180402-114759.pb --output model.onnx

    2020-08-03 20:18:05.081538: W tensorflow/stream_executor/platform/default/dso_loader.cc:59] Could not load dynamic library 'libcudart.so.10.1'; dlerror: libcudart.so.10.1: cannot open shared object file: No such file or directory
    2020-08-03 20:18:05.081680: I tensorflow/stream_executor/cuda/cudart_stub.cc:29] Ignore above cudart dlerror if you do not have a GPU set up on your machine.
    2020-08-03 20:18:07,431 - WARNING - '--tag' not specified for saved_model. Using --tag serve
    Traceback (most recent call last):
    File "/usr/lib/python3.8/runpy.py", line 193, in _run_module_as_main
    return _run_code(code, main_globals, None,
    File "/usr/lib/python3.8/runpy.py", line 86, in _run_code
    exec(code, run_globals)
    File "/usr/local/lib/python3.8/dist-packages/tf2onnx/convert.py", line 171, in
    main()
    File "/usr/local/lib/python3.8/dist-packages/tf2onnx/convert.py", line 131, in main
    graph_def, inputs, outputs = tf_loader.from_saved_model(
    File "/usr/local/lib/python3.8/dist-packages/tf2onnx/tf_loader.py", line 288, in from_saved_model
    _from_saved_model_v2(model_path, input_names, output_names, tag, signatures, concrete_function)
    File "/usr/local/lib/python3.8/dist-packages/tf2onnx/tf_loader.py", line 247, in _from_saved_model_v2
    imported = tf.saved_model.load(model_path, tags=tag) # pylint: disable=no-value-for-parameter
    File "/usr/local/lib/python3.8/dist-packages/tensorflow/python/saved_model/load.py", line 603, in load
    return load_internal(export_dir, tags, options)
    File "/usr/local/lib/python3.8/dist-packages/tensorflow/python/saved_model/load.py", line 614, in load_internal
    loader_impl.parse_saved_model_with_debug_info(export_dir))
    File "/usr/local/lib/python3.8/dist-packages/tensorflow/python/saved_model/loader_impl.py", line 56, in parse_saved_model_with_debug_info
    saved_model = _parse_saved_model(export_dir)
    File "/usr/local/lib/python3.8/dist-packages/tensorflow/python/saved_model/loader_impl.py", line 110, in parse_saved_model
    raise IOError("SavedModel file does not exist at: %s/{%s|%s}" %
    OSError: SavedModel file does not exist at: home/xesk/Desktop/2s/20180402-114759/20180402-114759.pb/{saved_model.pbtxt|saved_model.pb}

附加上下文

我没有运行任何 CUDA 或类似的东西，只有 CPU。 下载的model是20180402-114759 。 这是我第一次使用这些工具，而且我在这个 AI 世界中还是个初学者，所以我可能遗漏了一些明显的东西。 当然，我多次检查了路径和命令语法。 可能与我下载的文件格式有关？

编辑

根据Venkatesh Wadawadagi的回答，我选择了选项 1。更改.meta文件的名称解决了脚本无法识别的问题。

该脚本或多或少正确运行，并完成创建 export_dir 目录，其中export_dir > 0 > variables子文件夹。 然而，它们是空的。

控制台output是这样的：

xesk@xesk:~/Desktop/UP2S/ACROMEGALLY/20180402-114759$ python3 ./pb2sm
2020-08-10 16:02:26.128846: W tensorflow/stream_executor/platform/default/dso_loader.cc:55] Could not load dynamic library 'libcuda.so.1'; dlerror: libcuda.so.1: cannot open shared object file: No such file or directory
2020-08-10 16:02:26.129114: E tensorflow/stream_executor/cuda/cuda_driver.cc:318] failed call to cuInit: UNKNOWN ERROR (303)
2020-08-10 16:02:26.129137: I tensorflow/stream_executor/cuda/cuda_diagnostics.cc:156] kernel driver does not appear to be running on this host (xesk): /proc/driver/nvidia/version does not exist
2020-08-10 16:02:26.129501: I tensorflow/core/platform/cpu_feature_guard.cc:142] Your CPU supports instructions that this TensorFlow binary was not compiled to use: AVX2
2020-08-10 16:02:26.139076: I tensorflow/core/platform/profile_utils/cpu_utils.cc:94] CPU Frequency: 2592000000 Hz
2020-08-10 16:02:26.139506: I tensorflow/compiler/xla/service/service.cc:168] XLA service 0x44018d0 initialized for platform Host (this does not guarantee that XLA will be used). Devices:
2020-08-10 16:02:26.139520: I tensorflow/compiler/xla/service/service.cc:176]   StreamExecutor device (0): Host, Default Version
WARNING:tensorflow:From /usr/local/lib/python3.7/dist-packages/tensorflow_core/python/training/queue_runner_impl.py:391: QueueRunner.__init__ (from tensorflow.python.training.queue_runner_impl) is deprecated and will be removed in a future version.
Instructions for updating:
To construct input pipelines, use the `tf.data` module.
2020-08-10 16:02:32.681265: W tensorflow/core/framework/cpu_allocator_impl.cc:81] Allocation of 17676288 exceeds 10% of system memory.
Traceback (most recent call last):
  File "/usr/local/lib/python3.7/dist-packages/tensorflow_core/python/client/session.py", line 1365, in _do_call
    return fn(*args)
  File "/usr/local/lib/python3.7/dist-packages/tensorflow_core/python/client/session.py", line 1350, in _run_fn
    target_list, run_metadata)
  File "/usr/local/lib/python3.7/dist-packages/tensorflow_core/python/client/session.py", line 1443, in _call_tf_sessionrun
    run_metadata)
tensorflow.python.framework.errors_impl.FailedPreconditionError: Attempting to use uninitialized value InceptionResnetV1/Block8/Branch_0/Conv2d_1x1/BatchNorm/beta/Adam
     [[{{node save/SaveV2_1}}]]

During handling of the above exception, another exception occurred:

Traceback (most recent call last):
  File "./pb2sm", line 17, in <module>
    strip_default_attrs=True)
  File "/usr/local/lib/python3.7/dist-packages/tensorflow_core/python/util/deprecation.py", line 507, in new_func
    return func(*args, **kwargs)
  File "/usr/local/lib/python3.7/dist-packages/tensorflow_core/python/saved_model/builder_impl.py", line 595, in add_meta_graph_and_variables
    saver.save(sess, variables_path, write_meta_graph=False, write_state=False)
  File "/usr/local/lib/python3.7/dist-packages/tensorflow_core/python/training/saver.py", line 1193, in save
    raise exc
  File "/usr/local/lib/python3.7/dist-packages/tensorflow_core/python/training/saver.py", line 1176, in save
    {self.saver_def.filename_tensor_name: checkpoint_file})
  File "/usr/local/lib/python3.7/dist-packages/tensorflow_core/python/client/session.py", line 956, in run
    run_metadata_ptr)
  File "/usr/local/lib/python3.7/dist-packages/tensorflow_core/python/client/session.py", line 1180, in _run
    feed_dict_tensor, options, run_metadata)
  File "/usr/local/lib/python3.7/dist-packages/tensorflow_core/python/client/session.py", line 1359, in _do_run
    run_metadata)
  File "/usr/local/lib/python3.7/dist-packages/tensorflow_core/python/client/session.py", line 1384, in _do_call
    raise type(e)(node_def, op, message)
tensorflow.python.framework.errors_impl.FailedPreconditionError: Attempting to use uninitialized value InceptionResnetV1/Block8/Branch_0/Conv2d_1x1/BatchNorm/beta/Adam
     [[node save/SaveV2_1 (defined at /usr/local/lib/python3.7/dist-packages/tensorflow_core/python/framework/ops.py:1748) ]]

Original stack trace for 'save/SaveV2_1':
  File "./pb2sm", line 17, in <module>
    strip_default_attrs=True)
  File "/usr/local/lib/python3.7/dist-packages/tensorflow_core/python/util/deprecation.py", line 507, in new_func
    return func(*args, **kwargs)
  File "/usr/local/lib/python3.7/dist-packages/tensorflow_core/python/saved_model/builder_impl.py", line 589, in add_meta_graph_and_variables
    saver = self._maybe_create_saver(saver)
  File "/usr/local/lib/python3.7/dist-packages/tensorflow_core/python/saved_model/builder_impl.py", line 227, in _maybe_create_saver
    allow_empty=True)
  File "/usr/local/lib/python3.7/dist-packages/tensorflow_core/python/training/saver.py", line 828, in __init__
    self.build()
  File "/usr/local/lib/python3.7/dist-packages/tensorflow_core/python/training/saver.py", line 840, in build
    self._build(self._filename, build_save=True, build_restore=True)
  File "/usr/local/lib/python3.7/dist-packages/tensorflow_core/python/training/saver.py", line 878, in _build
    build_restore=build_restore)
  File "/usr/local/lib/python3.7/dist-packages/tensorflow_core/python/training/saver.py", line 499, in _build_internal
    save_tensor = self._AddShardedSaveOps(filename_tensor, per_device)
  File "/usr/local/lib/python3.7/dist-packages/tensorflow_core/python/training/saver.py", line 291, in _AddShardedSaveOps
    return self._AddShardedSaveOpsForV2(filename_tensor, per_device)
  File "/usr/local/lib/python3.7/dist-packages/tensorflow_core/python/training/saver.py", line 265, in _AddShardedSaveOpsForV2
    sharded_saves.append(self._AddSaveOps(sharded_filename, saveables))
  File "/usr/local/lib/python3.7/dist-packages/tensorflow_core/python/training/saver.py", line 206, in _AddSaveOps
    save = self.save_op(filename_tensor, saveables)
  File "/usr/local/lib/python3.7/dist-packages/tensorflow_core/python/training/saver.py", line 122, in save_op
    tensors)
  File "/usr/local/lib/python3.7/dist-packages/tensorflow_core/python/ops/gen_io_ops.py", line 1946, in save_v2
    name=name)
  File "/usr/local/lib/python3.7/dist-packages/tensorflow_core/python/framework/op_def_library.py", line 794, in _apply_op_helper
    op_def=op_def)
  File "/usr/local/lib/python3.7/dist-packages/tensorflow_core/python/util/deprecation.py", line 507, in new_func
    return func(*args, **kwargs)
  File "/usr/local/lib/python3.7/dist-packages/tensorflow_core/python/framework/ops.py", line 3357, in create_op
    attrs, op_def, compute_device)
  File "/usr/local/lib/python3.7/dist-packages/tensorflow_core/python/framework/ops.py", line 3426, in _create_op_internal
    op_def=op_def)
  File "/usr/local/lib/python3.7/dist-packages/tensorflow_core/python/framework/ops.py", line 1748, in __init__
    self._traceback = tf_stack.extract_stack()

有没有可能我缺少一些要安装的库？ 似乎与某些 CUDA 实现有关，但我没有。 可能吗？

Answer 1

您正在使用的命令：

python -m tf2onnx.convert --saved-model home/xesk/Desktop/2s/20180402-114759/20180402-114759.pb --output model.onnx

请注意，您正在使用的Fac.net trained model 只有冻结图（ .pb文件）和检查点（ .ckpt ）并且没有您的命令正在寻找的saved-model 。

所以基本上你将路径传递给冻结图的.pb文件，这与SavedModel的.pb文件（你没有）不同。 Savedmodel 将包含variables文件夹和saved_model.pb文件。

这就是错误的原因：

OSError: SavedModel file does not exist

在此处阅读有关 SavedModel 的更多信息。

要继续进行 ONNX 转换，您有两种选择：

1. 将检查点转换为 SavedModel：

为此使用以下代码：

import os
import tensorflow as tf

trained_checkpoint_prefix = 'model-20180402-114759.ckpt-275'
export_dir = os.path.join('export_dir', '0')

graph = tf.Graph()
with tf.compat.v1.Session(graph=graph) as sess:
    # Restore from checkpoint
    loader = tf.compat.v1.train.import_meta_graph(trained_checkpoint_prefix + '.meta')
    loader.restore(sess, trained_checkpoint_prefix)

    # Export checkpoint to SavedModel
    builder = tf.compat.v1.saved_model.builder.SavedModelBuilder(export_dir)
    builder.add_meta_graph_and_variables(sess,
                                         [tf.saved_model.TRAINING, tf.saved_model.SERVING],
                                         strip_default_attrs=True)
    builder.save() 

注意： .data 、 .index和.meta应该有相同的前缀，然后这段代码才能工作。 所以重命名.meta文件。

mv model-20180402-114759.meta model-20180402-114759.ckpt-275.meta

例如：

2. 使用ckpt文件或frozen-graph.pb进行 onnx 转换

从检查点格式：

python -m tf2onnx.convert --checkpoint tensorflow-model-meta-file-path --output model.onnx --inputs input0:0,input1:0 --outputs output0:0

来自 graphdef/frozen-graph 格式：

python -m tf2onnx.convert --graphdef tensorflow-model-graphdef-file --output model.onnx --inputs input0:0,input1:0 --outputs output0:0

如果您的 TensorFlow model 的格式不是saved model ，那么您需要提供 model 图的inputs和outputs 。

从这个：

如果您的 model 是检查点或graphdef格式，并且您不知道 model 的输入和 output 节点，则可以使用summarize_graph TensorFlow 实用程序。 summarize_graph工具确实需要从源代码下载和构建。 如果您可以选择前往您的 model 提供商并以保存的 model 格式获取 model，那么我们建议您这样做。

Answer 2

我遇到过类似的错误。 在我的例子中，我错误地给出了 pb 文件而不是path/to/savedmodel ，它应该是包含saved_model.pb的目录的路径。 因此，假设您的20180402-114759.pb位于目录home/xesk/Desktop/2s/20180402-114759中，命令应为：

python -m tf2onnx.convert --saved-model home/xesk/Desktop/2s/20180402-114759 --output model.onnx

有关详细信息，请参阅开始将 TensorFlow 转换为 ONNX和使用 SavedModel 格式。