What is model checkpointing in Keras?

Updated May 16, 2026

Short answer

It saves model weights during training.

Deep explanation

ModelCheckpoint callback stores best-performing model versions based on validation metrics.

Real-world example

Used in long training jobs on GPUs.

Common mistakes

  • Not saving best weights leading to suboptimal model.

Follow-up questions

  • What is save_best_only?
  • Where are checkpoints stored?

More Keras interview questions

View all →