From #52 , the hyperparameter tuner will use bad sampling periods when tuning against a re-sampled model. This patch corrects this. However, this is a result of counterintuitive interface design, and should likely be addressed by tweaking the class interfaces.
CC @Abdu-Hekal feel free to cherry-pick or merge. I ran it against your script for validation.
From #52 , the hyperparameter tuner will use bad sampling periods when tuning against a re-sampled model. This patch corrects this. However, this is a result of counterintuitive interface design, and should likely be addressed by tweaking the class interfaces.
CC @Abdu-Hekal feel free to cherry-pick or merge. I ran it against your script for validation.