Good Practice

Practical recommendations for running WeightsLab at scale — large datasets, long experiments, and production-like setups.

• Heavy-experiment loader flags
• Implementing get_items in your dataset class
• Training loop
• Signals and storage
• Summary table

Dataset and loaders — keeping a large dataset off the critical path: the array_* loader flags, and implementing get_items so label scans don’t pay for an image decode.

Training loop — why the loop should run until you stop it, and what a fixed step budget costs you.

Signals and storage — how much to send per step, and the three storage modes to choose between.