我训练的真人是美剧的一个人物，照片一共有52张，是白底人物抠图照，批次2次训练一次，4070显卡，你们觉得我应该怎么训练更像？而不过拟合。
steps: 0%| | 0/11700 [lbk]lbk[rbk]lbk[lbk]rbk[rbk]00:00<?, ?it/s[lbk]lbk[rbk]rbk[lbk]rbk[rbk]epoch 1/15steps: 7%|███▊ | 780/11700 [lbk]lbk[rbk]lbk[lbk]rbk[rbk]06:14<1:27:16, 2.09it/s, avr_loss=0.0885[lbk]lbk[rbk]rbk[lbk]rbk[rbk]saving checkpoint: ./output\ epoch 2/15steps: 13%|███████▌ | 1560/11700 [lbk]lbk[rbk]lbk[lbk]rbk[rbk]11:43<1:16:12, 2.22it/s, avr_loss=0.0934[lbk]lbk[rbk]rbk[lbk]rbk[rbk]saving checkpoint: ./output\epoch 3/15steps: 20%|███████████▍ | 2340/11700 [lbk]lbk[rbk]lbk[lbk]rbk[rbk]17:16<1:09:05, 2.26it/s, avr_loss=0.0874[lbk]lbk[rbk]rbk[lbk]rbk[rbk]saving checkpoint: ./output\epoch 4/15steps: 27%|███████████████▏ | 3120/11700 [lbk]lbk[rbk]lbk[lbk]rbk[rbk]22:50<1:02:47, 2.28it/s, avr_loss=0.0879[lbk]lbk[rbk]rbk[lbk]rbk[rbk]saving checkpoint: ./output\epoch 5/15steps: 33%|███████████████████▋ | 3900/11700 [lbk]lbk[rbk]lbk[lbk]rbk[rbk]28:24<56:49, 2.29it/s, avr_loss=0.0821[lbk]lbk[rbk]rbk[lbk]rbk[rbk]saving checkpoint: ./output\epoch 6/15steps: 40%|████████████████████████ | 4680/11700 [lbk]lbk[rbk]lbk[lbk]rbk[rbk]33:16<49:54, 2.34it/s, avr_loss=0.077[lbk]lbk[rbk]rbk[lbk]rbk[rbk]saving checkpoint: ./output\epoch 7/15steps: 47%|████████████████████████████ | 5460/11700 [lbk]lbk[rbk]lbk[lbk]rbk[rbk]37:50<43:15, 2.40it/s, avr_loss=0.071[lbk]lbk[rbk]rbk[lbk]rbk[rbk]saving checkpoint: ./output\epoch 8/15steps: 53%|███████████████████████████████▍ | 6240/11700 [lbk]lbk[rbk]lbk[lbk]rbk[rbk]42:24<37:06, 2.45it/s, avr_loss=0.0723[lbk]lbk[rbk]rbk[lbk]rbk[rbk]saving checkpoint: ./output\epoch 9/15steps: 60%|███████████████████████████████████▍ | 7020/11700 [lbk]lbk[rbk]lbk[lbk]rbk[rbk]47:01<31:20, 2.49it/s, avr_loss=0.0706[lbk]lbk[rbk]rbk[lbk]rbk[rbk]saving checkpoint: ./output\epoch 10/15steps: 67%|███████████████████████████████████████▎ |

epoch 10/15
steps: 67%|███████████████████████████████████████▎ | 7800/11700 [51:36<25:48, 2.52it/s, avr_loss=0.0641]
saving checkpoint: ./output
epoch 11/15
steps: 73%|███████████████████████████████████████████▎ | 8580/11700 [56:09<20:25, 2.55it/s, avr_loss=0.0592]
saving checkpoint: ./output
epoch 12/15
steps: 80%|█████████████████████████████████████████████▌ | 9360/11700 [1:00:42<15:10, 2.57it/s, avr_loss=0.0555]
saving checkpoint: ./output
epoch 13/15
steps: 87%|████████████████████████████████████████████████▌ | 10140/11700 [1:05:16<10:02, 2.59it/s, avr_loss=0.0523]
saving checkpoint: ./output
epoch 14/15
steps: 93%|████████████████████████████████████████████████████▎ | 10920/11700 [1:09:48<04:59, 2.61it/s, avr_loss=0.0478]
saving checkpoint: ./output
epoch 15/15
steps: 100%|████████████████████████████████████████████████████████| 11700/11700 [1:14:22<00:00, 2.62it/s, avr_loss=0.0445]
saving checkpoint: ./output\
2024-04-25 15:55:18 INFO model saved.
steps: 100%|████████████████████████████████████████████████████████| 11700/11700 [1:14:22<00:00, 2.62it/s, avr_loss=0.0445]
15:55:19-447518 INFO Training finished / 训练完成