Thanks to the author for your outstanding work and contribution to the open-source community. I would like to know what training settings the author used to achieve the experimental results listed in the table, including lr, augmentation_prob, epoch, decay_ratio, and decay_epoch. I want to obtain checkpoints with better performance, but I don’t have time to tune the parameters.
Thanks to the author for your outstanding work and contribution to the open-source community. I would like to know what training settings the author used to achieve the experimental results listed in the table, including lr, augmentation_prob, epoch, decay_ratio, and decay_epoch. I want to obtain checkpoints with better performance, but I don’t have time to tune the parameters.