Error running export_larq_model.py

[nathankopp]

I've trained QuickYOLOv2 on a custom dataset and now I'm trying to convert it to LCE/TFLite. Unfortunately, when I run python export_larq_model.py, I get the following error:

ValueError: Input 0 of node model_1/batch_normalization_22/AssignNewValue was passed float from model_1/batch_normalization_22/FusedBatchNormV3/ReadVariableOp/resource:0 incompatible with expected resource.

Here's the full stack trace:

Traceback (most recent call last):
  File "C:\Users\x\AppData\Roaming\Python\Python38\site-packages\tensorflow\python\framework\importer.py", line 496, in _import_graph_def_internal
    results = c_api.TF_GraphImportGraphDefWithResults(
tensorflow.python.framework.errors_impl.InvalidArgumentError: Input 0 of node model_1/batch_normalization_22/AssignNewValue was passed float from model_1/batch_normalization_22/FusedBatchNormV3/ReadVariableOp/resource:0 incompatible with expected resource.

During handling of the above exception, another exception occurred:

Traceback (most recent call last):
  File "export_larq_model.py", line 124, in <module>
    flatbuffer_bytes = lce.convert_keras_model(yolo)
  File "C:\Users\x\AppData\Local\Programs\Python\Python38\lib\site-packages\larq_compute_engine\mlir\python\converter.py", line 118, in convert_keras_model
    frozen_func = convert_variables_to_constants_v2(func, lower_control_flow=False)
  File "C:\Users\x\AppData\Roaming\Python\Python38\site-packages\tensorflow\python\framework\convert_to_constants.py", line 1076, in convert_variables_to_constants_v2
    return _construct_concrete_function(func, output_graph_def,
  File "C:\Users\x\AppData\Roaming\Python\Python38\site-packages\tensorflow\python\framework\convert_to_constants.py", line 1001, in _construct_concrete_function
    new_func = wrap_function.function_from_graph_def(output_graph_def,
  File "C:\Users\x\AppData\Roaming\Python\Python38\site-packages\tensorflow\python\eager\wrap_function.py", line 650, in function_from_graph_def
    wrapped_import = wrap_function(_imports_graph_def, [])
  File "C:\Users\x\AppData\Roaming\Python\Python38\site-packages\tensorflow\python\eager\wrap_function.py", line 621, in wrap_function
    func_graph.func_graph_from_py_func(
  File "C:\Users\x\AppData\Roaming\Python\Python38\site-packages\tensorflow\python\framework\func_graph.py", line 990, in func_graph_from_py_func
    func_outputs = python_func(*func_args, **func_kwargs)
  File "C:\Users\x\AppData\Roaming\Python\Python38\site-packages\tensorflow\python\eager\wrap_function.py", line 87, in __call__
    return self.call_with_variable_creator_scope(self._fn)(*args, **kwargs)
  File "C:\Users\x\AppData\Roaming\Python\Python38\site-packages\tensorflow\python\eager\wrap_function.py", line 93, in wrapped
    return fn(*args, **kwargs)
  File "C:\Users\x\AppData\Roaming\Python\Python38\site-packages\tensorflow\python\eager\wrap_function.py", line 648, in _imports_graph_def
    importer.import_graph_def(graph_def, name="")
  File "C:\Users\x\AppData\Roaming\Python\Python38\site-packages\tensorflow\python\util\deprecation.py", line 538, in new_func
    return func(*args, **kwargs)
  File "C:\Users\x\AppData\Roaming\Python\Python38\site-packages\tensorflow\python\framework\importer.py", line 400, in import_graph_def
    return _import_graph_def_internal(
  File "C:\Users\x\AppData\Roaming\Python\Python38\site-packages\tensorflow\python\framework\importer.py", line 501, in _import_graph_def_internal
    raise ValueError(str(e))
ValueError: Input 0 of node model_1/batch_normalization_22/AssignNewValue was passed float from model_1/batch_normalization_22/FusedBatchNormV3/ReadVariableOp/resource:0 incompatible with expected resource.

[tehtea]

As for this issue, I wasn't able to replicate it locally... Will try again on Google Colab. These are your environment settings right?

quickyolov2 model
CUDA 11.2
CUDNN 8.1.1.33
Tensorflow 2.4.1
Larq 0.11.2
LCE 0.5.0

[nathankopp]

Yes, those are my environment settings. I have found a solution. Larq requires Tensorflow 2.3.0 and does not work properly with version 2.4.1.

Using pip install tensorflow==2.3.0 solved my problem.