Open Models Almanac
Methodsoptimizationtraining runtime

single mention · optimization

Frequent checkpointing for fault tolerance

Saves checkpoints frequently as part of a fault-tolerance setup, limiting lost work after a failure.

The sources on this page only mention it in passing. The line above is the extraction's one-line gloss. See training runtime for the methods filed beside it.

Also called frequent checkpointing.

Filed under optimization › training runtime