unable to autofill for 'bad_output_shape', model tensor configurations are contradicting each other in terms of whether batching is supported