hi, i am extracting the weights but weights are in floating point. I am using conversion if weights are negative taking it as -1 and if positive taking it as +1. But accuracy is dropping with these. First test accuracy was 92.94% and then it dropped to 28.1%
Environment
TensorFlow version:
Larq version:
hi, i am extracting the weights but weights are in floating point. I am using conversion if weights are negative taking it as -1 and if positive taking it as +1. But accuracy is dropping with these. First test accuracy was 92.94% and then it dropped to 28.1%
Environment
TensorFlow version:
Larq version: