【AI达人养成营】炼丹trick
收藏
1. DropOut
2. ReLu
3. Normalization
4. Skip connection:
5. big batch size
6. Consie Learning rate Descay:
7. Label smoothing
8. Focal loss
9. warm up
10. EMA model
11. lr_scheduler1. DropOut
2. ReLu
3. Normalization
4. Skip connection:
5. big batch size
6. Consie Learning rate Descay:
7. Label smoothing
8. Focal loss
9. warm up
10. EMA model
11. lr_scheduler
0
收藏
请登录后评论
