KeyError: "The name 'deepspeech/logits:0' refers to a Tensor which does not exist. The operation, 'deepspeech/logits', does not exist in the graph."

[Hybrogen]

error:

~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~ [] ~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~
Traceback (most recent call last):
  File "D:\HWork\NEW08\AD-NeRF\data_util\deepspeech_features\extract_ds_features.py", line 128, in <module>
    main()
  File "D:\HWork\NEW08\AD-NeRF\data_util\deepspeech_features\extract_ds_features.py", line 120, in main
    extract_features(
  File "D:\HWork\NEW08\AD-NeRF\data_util\deepspeech_features\extract_ds_features.py", line 80, in extract_features
    conv_audios_to_deepspeech(
  File "D:\HWork\NEW08\AD-NeRF\data_util\deepspeech_features\deepspeech_features.py", line 41, in conv_audios_to_deepspeech
    graph, logits_ph, input_node_ph, input_lengths_ph = prepare_deepspeech_net(
  File "D:\HWork\NEW08\AD-NeRF\data_util\deepspeech_features\deepspeech_features.py", line 106, in prepare_deepspeech_net
    logits_ph = graph.get_tensor_by_name("deepspeech/logits:0")
  File "D:\HPrograms\H_OnInstall\Python\Python310\lib\site-packages\tensorflow\python\framework\ops.py", line 4188, in get_tensor_by_name
    return self.as_graph_element(name, allow_tensor=True, allow_operation=False)
  File "D:\HPrograms\H_OnInstall\Python\Python310\lib\site-packages\tensorflow\python\framework\ops.py", line 4012, in as_graph_element
    return self._as_graph_element_locked(obj, allow_tensor, allow_operation)
  File "D:\HPrograms\H_OnInstall\Python\Python310\lib\site-packages\tensorflow\python\framework\ops.py", line 4052, in _as_graph_element_locked
    raise KeyError("The name %s refers to a Tensor which does not "
KeyError: "The name 'deepspeech/logits:0' refers to a Tensor which does not exist. The operation, 'deepspeech/logits', does not exist in the graph."

i find this error in ops.py and find out this:

# line 4049
if op_name in self._nodes_by_name:
  op = self._nodes_by_name[op_name]
else:
  raise KeyError("The name %s refers to a Tensor which does not "
                 "exist. The operation, %s, does not exist in the "
                 "graph." % (repr(name), repr(op_name)))

so i try to print [tensor.name for tensor in graph._nodes_by_name] before logits_ph = graph.get_tensor_by_name("deepspeech/logits:0") in deepspeech_features.py.
but it is empty. i wonder why and how to fix it~

[Hybrogen]

find out reason is that the version of tensorflow which origin project used is 1.15.2, but new version has some different.
anyway this change will be solve:

# data_util/deepspeech_features/deepspeech_features.py
def prepare_deepspeech_net(deepspeech_pb_path):
    # Load graph and place_holders:
    with tf.io.gfile.GFile(deepspeech_pb_path, "rb") as f:
        graph_def = tf.compat.v1.GraphDef()
        graph_def.ParseFromString(f.read())

    graph = tf.compat.v1.get_default_graph()

    tf.import_graph_def(graph_def, name="deepspeech")
    # check all graphs
    # print('~'*50, [tensor for tensor in graph._nodes_by_name], '~'*50)
    # print('~'*50, [tensor.name for tensor in graph.get_operations()], '~'*50)
    # i modified
    logits_ph = graph.get_tensor_by_name("logits:0")
    input_node_ph = graph.get_tensor_by_name("input_node:0")
    input_lengths_ph = graph.get_tensor_by_name("input_lengths:0")
    # original
    # logits_ph = graph.get_tensor_by_name("deepspeech/logits:0")
    # input_node_ph = graph.get_tensor_by_name("deepspeech/input_node:0")
    # input_lengths_ph = graph.get_tensor_by_name("deepspeech/input_lengths:0")

    return graph, logits_ph, input_node_ph, input_lengths_ph