aaa runner = dict(type=’EpochBasedRunner’, max_epochs=36) evaluation = dict(interval=4, metric=’bbox’, save_best=’bbox_mAP’) auto_scale_lr = dict(enable=False, base_batch_size=16) 10 * 36 = 2.5 * (4*36) step 144 epoch 전체 데이터 셋에 대해 1번 forward pass/backward pass 과정(Back propagation algorithm, Read more…