if it is neccessry, change the fix batch input to dynamic batch input, and verify the fp32 and int8 model have the same output.
onnx_model_path = 'resnet50_3x224x224_op12.onnx'
qt_onnx_model_path = 'resnet50-v1-12-int8/resnet50-v1-12-int8_dynamic.onnx'
convert_model_batch_to_dynamic('resnet50-v1-12-int8/resnet50-v1-12-int8.onnx')
verify_quantized_model(onnx_model_path, qt_onnx_model_path)
extract_node_info(model_path=qt_onnx_model_path,
input_shapes = {input:[1,3,224,224]},
dynamic_input_shape=True
)