Skip to content

Release 0.3.7

Compare
Choose a tag to compare
@workingloong workingloong released this 13 May 06:02
· 492 commits to master since this release

Features:

  • Flash Checkpoint suppors deleting old checkpoints.

BugFix:

  • Save/load the non-params-related variables of dist optimizer in Megatron-LM models.
  • The agent waits for async saving checkpoint finishes before exiting.