Training for magic_wand not working

TensorFlow General Discussion
, ,
1

Hi, people.

I am trying to train the model for the magic_wand example with the notebook and the given dataset from the repository(https://github.com/tensorflow/tflite-micro/blob/main/tensorflow/lite/micro/examples/magic_wand/train/train_magic_wand_model.ipynb). But when I change the data model to the one from the notebook in ‘magic_wand_model_data.cpp’ and upload the program to the Nano Sense from the Arduino IDE, the board is no longer recognised by the PC and I have to reset it (RST button just after connecting the board).
Running the program with the default data model everything works perfectly.

Best regards. Thank you.

3

I have finally found the error. In theory this error should be reported by the IDE, but in my case it was not. For those of you who are in my situation, you have to edit the .ino of your project and add in the line 86:

micro_mutable_op_resolver.AddBuiltin(tflite::BuiltinOperator_RESHAPE, tflite::ops::micro::Register_RESHAPE());