Sagemaker Neo compilation failed for TF Lite object detection model

I am trying to compile a TFLite object detection model for imx8qm using Sagemaker Neo. I supplied a .tflite file and specified the input config as `{"input":[1, 300, 300, 3]}`
But the compilation failed with the following error:
`ClientError: InputConfiguration: Framework cannot load TFLite model. Unable to infer tensor data type for all inputs/outputs: Please specify all input layers in data_shape.`

How to fix this?

Start with the supplied .tflite model and the SageMaker Neo InputConfiguration for the imx8qm target. Re-run compilation while comparing the model's input layers with the declared data_shape, and consider the issue complete when the model compiles without the reported tensor data type error.

Written by the indexing model from the issue text.