I would like to ask, if I want to convert two tflites into full integers (including input and output with int8), then I know the calibration data of the first tflite. So, should I use the value generated by the first tflite that has already been quantized with int8 and inferred, or regenerate it into a tflite that does not undergo integer quantization and generates the value after inference as the calibration data for the second tflite. Or how does this calibration data need to be generated？
Hello:
I would like to ask, if I want to convert two tflites into full integers (including input and output with int8), then I know the calibration data of the first tflite. So, should I use the value generated by the first tflite that has already been quantized with int8 and inferred, or regenerate it into a tflite that does not undergo integer quantization and generates the value after inference as the calibration data for the second tflite. Or how does this calibration data need to be generated？
Thanks!