Save checkpoints with accelerator

[yujianll]

When using accelerator for training, how can we only save the additional lora weights for temporary checkpoints? For example, at this line, how to avoid saving the whole model?

[github-actions[bot]]

This issue has been automatically marked as stale because it has not had recent activity. If you think this still needs to be addressed please comment on this thread.

[samiur]

Hi, this is still an issue. Please reopen the issue.

[kylin-zhou]

I have same problem, I use deepspeed, help !