why train so much epoch(240epoch) than caffe ssd (12000iter==58epoch)

[yinglang]

for origin caffe ssd, only train 120000 iter, batch size=8, take only about 58 epochs, why you train your ssd in mxnet need 240epoch, it cost too much.
do you try 58 epoch train use decay in epoch 39 and epoch 49?
it very confuse me, thanks for you read.

[zhreshold]

the original paper take batch size =32, so that's 240 epoch equivalent.