samples/python/tensorflow_object_detection_api/build_engine.py fails for models with dynamic axes

[fdarvas]

I tried to use the model conversion script samples/python/tensorflow_object_detection_api/build_engine.py to build and engine from an onnx model. I have confirmed that the onnx model can be used by trtexec, which correctly recognized the dynamic axes, but the python script fails with

python3 build_engine.py --onnx myModel.onnx --engine myModel.engine --precision fp32

INFO:EngineBuilder:Network Description
INFO:EngineBuilder:Input 'images' with shape (-1, 3, 640, 640) and dtype DataType.FLOAT
INFO:EngineBuilder:Input 'orig_target_sizes' with shape (-1, 2) and dtype DataType.INT64
INFO:EngineBuilder:Output 'classes_int64' with shape (-1, 300) and dtype DataType.INT64
INFO:EngineBuilder:Output 'boxes' with shape (-1, 300, 4) and dtype DataType.FLOAT
INFO:EngineBuilder:Output 'scores' with shape (-1, 300) and dtype DataType.FLOAT
INFO:EngineBuilder:Output 'classes' with shape (-1, 300) and dtype DataType.FLOAT
Traceback (most recent call last):
  File "/mnt/e/data/teknoir/TensorRT/samples/python/tensorflow_object_detection_api/build_engine.py", line 334, in <module>
    main(args)
  File "/mnt/e/data/teknoir/TensorRT/samples/python/tensorflow_object_detection_api/build_engine.py", line 269, in main
    builder.create_network(args.onnx)
  File "/mnt/e/data/teknoir/TensorRT/samples/python/tensorflow_object_detection_api/build_engine.py", line 182, in create_network
    assert self.batch_size > 0
AssertionError

This is obviously a built-in feature, but there are many models out there with dynamic axes - for trtexec I can speficy a min, opt and max range - but there are no such options for this example?