[Python] Tensorflow lite sample code does not work from the tutorials

https://ai.google.dev/edge/litert/models/signatures#java I have the model generated, I have the model loaded. I verified this because I can see the signatures that are available:

I don't know what the encoded_result and decoded_result variables should be. Can someone give me a concrete example code?

val encodeInput = mutableMapOf<String, Any>()
val encodeOutput = mutableMapOf<String, Any>()

encodeInput["x"] = arrayOf(floatArrayOf(9f))
encodeOutput["encoded_result"] = arrayOf(arrayOf("9"))

interpreter!!.runSignature(encodeInput,encodeOutput, "encode")

val decodeInput = mutableMapOf<String, Any>()
val decodeOutput = mutableMapOf<String, Any>()

decodeOutput["decoded_result"] = arrayOf(arrayOf("9"))
decodeInput["x"] = arrayOf(floatArrayOf(9f))

interpreter!!.runSignature(decodeInput, decodeOutput, "decode")`


This is what I have and this is the tf functions:

`

@tf.function(input_signature=[tf.TensorSpec(shape=[None], dtype=tf.float32)])
def encode(self, x):
result = tf.strings.as_string(x)
return {
"encoded_result": result
}

@tf.function(input_signature=[tf.TensorSpec(shape=[None], dtype=tf.string)])
def decode(self, x):
result = tf.strings.to_number(x)
return {
"decoded_result": result
}`


I understand the encodeInput[x] we're passing in the value of x to be encoded but are we also telling the function what the output of x will be ? encoded_result: result? If so , what is the point of the method if we tell it what the input and output will be for both the encode and decode function? If I can working kotlin code that would help me IMMENSELY. Thank you!

I've tried countless example code and I expected no errors and the decodeOutput to be of the format encoded_result: "9" "decoded_result: 9

but I keep getting these Cannot convert between a TensorFlowLite tensor with type STRING and a Java object of type [[F (which is compatible with the TensorFlowLite type FLOAT32).

and also changed code around and was getting: StringToNumberOp errors

things I have tried: I was able to just create a simple addition function to take in a number and add 5 to it and that works but when i try to use string manipulations that fails. I want to be able to pass in lists of string data and have that be fed to the model for training on device specifically.

Continue reading...