Hello
I have a keras model, trained in channel_first format. I converted the model to tflite, but it adds an additional transpose layer to the tflite model. My target board doesn’t support transpose layer, is there any other way to carry out the conversion. The model has to be trained on channel_first format only.