DeepSpeed
Separate ZeRO3 InflightParamRegistry for train and eval
#3884
Merged

Separate ZeRO3 InflightParamRegistry for train and eval #3884

HeyangQin merged 9 commits into master from HeyangQin/fix_pr_3462
HeyangQin
awan-10 extend the test and fix fp16 typo.
17228f43
awan-10 guard reset params with z3 enabled check.
709bbdd4
awan-10 Merge branch 'master' into fix-he-lora-z3-test
cbbba7c7
HeyangQin create standalone registries for training and eval respectively
badfb23a
HeyangQin HeyangQin requested a review from jeffra jeffra 2 years ago
HeyangQin HeyangQin requested a review from tjruwase tjruwase 2 years ago
HeyangQin HeyangQin requested a review from samyam samyam 2 years ago
HeyangQin HeyangQin requested a review from mrwyattii mrwyattii 2 years ago
HeyangQin make this pr standalone
1fcbd844
HeyangQin standalone v2
f270a2b4
HeyangQin HeyangQin requested a review from RezaYazdaniAminabadi RezaYazdaniAminabadi 2 years ago
HeyangQin HeyangQin requested a review from awan-10 awan-10 2 years ago
HeyangQin HeyangQin requested a review from cmikeh2 cmikeh2 2 years ago
HeyangQin HeyangQin requested a review from arashb arashb 2 years ago
HeyangQin Merge branch 'master' into HeyangQin/fix_pr_3462
419461b8
HeyangQin HeyangQin enabled auto-merge (squash) 2 years ago
awan-10
awan-10 approved these changes on 2023-07-05
HeyangQin Merge branch 'master' into HeyangQin/fix_pr_3462
ec0d9eb2
mrwyattii Merge branch 'master' into HeyangQin/fix_pr_3462
cd2ba62f
HeyangQin HeyangQin merged 9377921a into master 2 years ago
mrwyattii mrwyattii deleted the HeyangQin/fix_pr_3462 branch 2 years ago
shyustc

Login to write a write a comment.

Login via GitHub

Assignees
No one assigned
Labels
Milestone