[fix auto-microbatch] FSDP reshard and cleanup after OOM to fix the cuda memory leak#3030
Merged
bigning merged 22 commits intodevfrom reshard-after-oomFeb 22, 2024
+129-2
Commits
Commits on Feb 18, 2024
- committed
- committed
- committed
- committed
Commits on Feb 21, 2024
- committed
- authored
- committed
- committed
- committed
- authored
- committed
- committed
- committed
