Saving optimizer state
Open
enhancement
- Dominant language
- C++
- Stars
- 3.4k
- Forks
- 701
- PR merge metrics
- No merged PRs in 30d
Description
Currently there are no methods to save the state of the optimizer, for instance current learning rate or history of gradients. These may be useful when resuming training from a model snapshot.
Could we have an API for this? I think a nice interface would be (for Python) `trainer.populate` after creating the Trainer object given the Model.
Contributor guide
No contributing guide indexed for this repository
Assessment
This issue has not been assessed yet.