Tingquan Gao
28ef1c00d8
[cherry-pick] some fix ( #2920 )
...
* fix model name
* fix: bug when distillation
* modify some default hyperparams to adapt to fine-tune downstream tasks
1. unset EMA because of the relatively small size of most downstream dataset;
2. use mean and std of IMN.
2023-08-25 22:02:57 +08:00
Tingquan Gao
7243f1429b
Revert "rm codes for compatibility with old version"
...
This reverts commit 6e77bd6cd5
.
2023-03-14 16:47:13 +08:00
gaotingquan
6e77bd6cd5
rm codes for compatibility with old version
2023-03-10 16:56:55 +08:00
cuicheng01
4924749e25
fix_celoss
2022-11-14 21:34:52 +08:00
dongshuilong
bcecbd1159
fix celoss bug
2022-10-27 14:17:09 +08:00
dongshuilong
944763d7a5
add fixmatch
2022-10-25 12:04:22 +08:00
gaotingquan
ba2dd01a13
refactor: deprecate MixCELoss
2021-10-27 19:47:43 +08:00
littletomatodonkey
85e2407e46
fix mix training for static program ( #1234 )
2021-09-15 14:36:11 +08:00
littletomatodonkey
81fcc9cb72
fix label smooth bug ( #1191 )
2021-09-01 19:36:28 +08:00
cuicheng01
592f5b10f5
support batch mix
2021-07-13 06:38:01 +00:00
littletomatodonkey
422dd57c38
fix ce loss with label smooth ( #838 )
2021-06-13 00:59:15 +08:00
littletomatodonkey
3b4f5f4dfc
add distillation and fix some apis ( #810 )
...
* fix save load and imagenet dataset
* refine trainer
2021-06-10 16:30:05 +08:00
weishengyu
69f563d234
rename losses -> loss
2021-06-03 15:17:49 +08:00