Closed mo-sameh closed 5 months ago
Previously, MS2 predictions were reordered after each test epoch to calculate the metrics because PeptDeep changes the order in place during prediction, which for larger datasets takes more time than the prediction itself. Now, we reorder the model input only once before training to match the order used in PeptDeep.
Is it better to change APD's method?
I think the best would be to support both:
SpecLibBase.refine_xxx()
Previously, MS2 predictions were reordered after each test epoch to calculate the metrics because PeptDeep changes the order in place during prediction, which for larger datasets takes more time than the prediction itself. Now, we reorder the model input only once before training to match the order used in PeptDeep.