MLCommons Algorithmic Efficiency is a benchmark and competition measuring neural network training speedups due to algorithmic improvements in both training algorithms and models.
Currently --save_checkpoints flag is only used to control saving intermediate checkpoints.
Our checkpointing code currently is not compatible with all possible submissions, so we will need a flag to completely disable checkpointing.
We should also clarify in the technical documentation that during scoring we will not use checkpoints.
Description
Currently --save_checkpoints flag is only used to control saving intermediate checkpoints. Our checkpointing code currently is not compatible with all possible submissions, so we will need a flag to completely disable checkpointing.
We should also clarify in the technical documentation that during scoring we will not use checkpoints.