pytc.vmc.mcmc_utils.save_walkers¶

pytc.vmc.mcmc_utils.save_walkers(walkers, filepath)[source]¶

Save a Walker’s full state (positions plus cached psi/det/grad/lap fields) to an HDF5 file, for continuing MCMC sampling in a later, separate process invocation without re-burning-in.

Distinct from save_optimization_history: this is walker state, not a training trajectory, so it has no ‘step’ axis and no cost/energy log.

Parameters:

walkers – A pytc.vmc.walker.Walker instance.
filepath (str) – Path to the HDF5 file (e.g., ‘walkers_checkpoint.h5’).

Returns:

Path to the saved HDF5 file.

Return type:

str