[fsdp, megatron] Refactor fully-async training to support multiple checkpoint engine backends#5029
Draft
Shangwei-Li wants to merge 1 commit intoverl-project:mainfrom
Draft
[fsdp, megatron] Refactor fully-async training to support multiple checkpoint engine backends#5029Shangwei-Li wants to merge 1 commit intoverl-project:mainfrom
Shangwei-Li wants to merge 1 commit intoverl-project:mainfrom