Closed SaiAditya2595 closed 2 years ago
These, among other hyper-parameters, are often chosen using cross-validation on a held out data set. The model complexity increases with the number of trained calibration keypoints, so there might be overfitting if the number is too large for the size of the training dataset.
I have seen some TF lattice files where pwl_calibration_num_keypoints are 5,10,20. How to select and set these keypoints?