n_dow = 7 # number of division for dayofweek/dayofmonth/hourofday n_dom = 31 n_hour = 24 n_clients = 57124 #57105 n_stands = 63 n_begin_end_pts = 5 # how many points we consider at the beginning and end of the known trajectory n_end_pts = 5 train_size = 1710670 n_valid = 1000 dim_embed = 50 dim_input = n_begin_end_pts * 2 * 2 + dim_embed + dim_embed dim_hidden = [200] dim_output = 2 learning_rate = 0.002 momentum = 0.9 batch_size = 32