nvalchemi-dynamics-hooks

You can monitor energy drift by registering the prebuilt EnergyDriftMonitorHook into your dynamics workflow. This callback safely observes simulation steps and logs deviations without modifying the core engine.

The best way is attaching prebuilt LoggingHook and ProfilerHook classes to your dynamics stages. These hooks execute stage-aware callbacks to capture timing data and log metrics throughout the simulation pipeline.

Yes, you can detect NaN values automatically by registering the NaNDetectorHook. It operates as a stage-aware callback within the dynamics workflow to enforce safety constraints and identify invalid states during simulation steps.

Create a custom hook class implementing the compatible Hook protocol and attach it to your Dynamics workflow. The simple registration API allows you to observe, log, and adjust batch state across each simulation step.

You need to snapshot batches during dynamics workflows when debugging reproducible instrumentation or validating intermediate states. The prebuilt SnapshotHook captures batch states at specific stages without altering core simulation engines.

Yes, the hooks system works with long-running simulations by providing extensible stage-aware execution. You can register callbacks to monitor energy drift, log metrics, and enforce safety constraints across dynamic steps without changing core logic.