|
|
543beefa2a
|
删除残差路径和shortcut，镜像问题仍存在
|
2026-01-16 15:21:47 +08:00 |
|
|
|
a92a0b29e9
|
更新模型结构，大步长反卷积后移，启用BN和tanh
|
2026-01-15 21:12:27 +08:00 |
|
|
|
df703638da
|
清理代码，删除跳连接部分
|
2026-01-11 13:25:34 +08:00 |
|
|
|
12de74f130
|
完善了跳连接，在上decode块后增加特征精炼层，未测效果
|
2026-01-09 18:23:45 +08:00 |
|
|
|
f7601e9170
|
初步可跑通，但loss计算有问题，不收敛
|
2026-01-08 09:43:23 +08:00 |
|
|
|
7e9564ef20
|
test modify swiftformer to temporal input
|
2026-01-07 11:03:33 +08:00 |
|
caitianren
|
0ddadad723
|
Fix this bug when setting distillation-type to none
|
2023-11-29 20:15:00 +08:00 |
|
Abdelrahman Shaker
|
37a4fe953d
|
Update swiftformer.py
|
2023-07-26 01:32:27 +04:00 |
|
amshaker
|
670dea3e1f
|
update README.md
|
2023-07-26 00:59:51 +04:00 |
|
amshaker
|
574907f49b
|
Initial release of SwiftFormer
|
2023-03-26 23:31:59 +04:00 |
|
