Commit Graph

16 Commits (1986116527b10ae348711be2b6a7daff136adf82)

Author SHA1 Message Date
Ryan (张磊) aa7645a831
Merge pull request #4 from blankde/feature_add_moe_refactor_zl
refactor code
2023-09-22 14:22:45 +08:00
Wenwen Qu 9e6e7986b6 refactor code for log 2023-09-22 14:14:58 +08:00
zhanglei 80972ff314 refactor code 2023-09-22 11:47:05 +08:00
Qu Wenwen 4a47872382 refactor code 2023-09-19 12:30:40 +08:00
zhanglei edc18bcddd fix precision inconsistency 2023-09-18 20:54:52 +08:00
Qu Wenwen 5aa5c96ec8 refactor split_moe_group code 2023-09-15 16:55:16 +08:00
Wenwen Qu 6cf0fec314 replace flashatten experts by feedforward experts 2023-09-08 18:04:57 +08:00
Wenwen Qu b021995199 fix bugs 2023-08-30 16:14:33 +08:00
Wenwen Qu f3da80a7ca reformat code 2023-08-28 14:46:03 +08:00
Wenwen Qu 0e6b1f856c add support for moe checkpoint 2023-08-24 17:01:14 +08:00
Wenwen Qu 929dd36cf2 avoid moe parameter partition in zero optimizer 2023-08-15 17:52:51 +08:00
Wenwen Qu 5f2e082b21 reformat code 2023-08-09 16:03:47 +08:00
zhanglei cdf3ed9533 add residual and other moe features 2023-08-09 14:14:18 +08:00
Wenwen Qu 9c8b999291 modified: internlm/model/moe.py
modified:   internlm/moe/sharded_moe.py
	modified:   internlm/utils/parallel.py
2023-08-08 16:46:14 +08:00
Wenwen Qu 8b198b2665 modified: .pre-commit-config.yaml
modified:   internlm/model/moe.py
	modified:   internlm/model/modeling_internlm.py
2023-08-08 15:44:37 +08:00
Wenwen Qu c357288a8b feat(XXX): add moe 2023-08-07 20:17:49 +08:00