I'm sorry to bother you that, here is the code:
min_level_weight, max_level_weight = int(cur_level+1)-cur_level, cur_level-int(cur_level)
I wonder if the 'cur_level' need to be changed during the iterations ? And when the iteration grows bigger the max_level_weight grows bigger too
But in the code It has be set as follows and never change
I'm sorry to bother you that, here is the code: min_level_weight, max_level_weight = int(cur_level+1)-cur_level, cur_level-int(cur_level)
I wonder if the 'cur_level' need to be changed during the iterations ? And when the iteration grows bigger the max_level_weight grows bigger too But in the code It has be set as follows and never change
phases = {'stabilize':[0, train_kimg//batch_size], 'fade_in':[train_kimg//batch_size+1, (transition_kimg+train_kimg)//batch_size]}