onnx autocast raise error with dynamic input shape

[lzcchl]

I think this is no good, because now a lot of model is dynamic input shape, I think it shoulds be a warning rather than a error. (ValueError: Input shape from 'images' does not match provided input shape: [0, 3, 0, 0] vs [1, 3, 480, 640]. Please make sure that your calibration data matches the ONNX input shapes.)

https://github.com/NVIDIA/Model-Optimizer/blob/main/modelopt/onnx/autocast/referencerunner.py#L93
raise ValueError ==>logger.warning

will it changed future release?