Evaluation

Evaluation#

Configuration reference for jaqmc solid evaluate. This page shows the effective defaults for the evaluation workflow preset. Use --dry-run to see the resolved config for your run, or add workflow.config.verbose=true to include field descriptions. Evaluation has only one stage, so stage keys (run.*, sampler.*, writers.*) live at the config root rather than under a train.* prefix. Defaults are resolved in this order: schema defaults, workflow preset, YAML config, then CLI overrides. For training config, see Training.

Workflow (workflow.*)#

These keys control evaluation-wide settings and checkpoint loading.

workflow.seed

Default: None · Type: int | None

Fixed random seed.

workflow.batch_size

Default: 4096 · Type: int

Number of walkers (samples) to use in each iteration.

workflow.save_path

Default: '' · Type: str

Path to save checkpoints and logs.

workflow.restore_path

Default: '' · Type: str

Path to restore checkpoints from.

workflow.config.ignore_extra

Default: False · Type: bool

If True, silently ignore unrecognized config keys.

workflow.config.verbose

Default: False · Type: bool

If True, print the fully resolved config with field descriptions at startup.

workflow.disable_jit

Default: False · Type: bool

Disable JAX JIT compilation (for debugging).

workflow.source_path

Default: (required) · Type: str

Path to the training run directory or checkpoint file to load parameters from.

System (system.*)#

Must match the training run. The effective defaults and built-in module choices are identical to the training system config.

Wavefunction (wf.*)#

Must match the training run. The effective defaults and built-in module choices are identical to the training wavefunction config.

Run Options (run.*)#

Evaluation reuses the same checkpointing and sampling controls as training, but adds digest_step_interval for previewing accumulated statistics.

run.check_vma

Default: True · Type: bool

Enable JAX validity checks during shard_map.

run.iterations

Default: 100 · Type: int

Total number of iterations to run.

run.burn_in

Default: 0 · Type: int

Sampling iterations to discard before the main loop for MCMC equilibration.

run.save_time_interval

Default: 600 · Type: int

Minimum wall-clock seconds between checkpoint saves.

run.save_step_interval

Default: 1000 · Type: int

Save checkpoints only at steps that are multiples of this value.

run.digest_step_interval

Default: 0 · Type: int

Log a preview of the accumulated evaluation digest every this many steps.

Sampler (sampler.*)#

  • Default sampler module: mcmc, and its effective keys are listed below.

sampler.steps

Default: 10 · Type: int

Number of Metropolis-Hastings updates per sample draw.

sampler.initial_width

Default: 0.1 · Type: float

Initial width (stddev) of the Gaussian proposal.

sampler.adapt_frequency

Default: 100 · Type: int

Frequency of adaptive width updates.

sampler.pmove_range

Default: (0.5, 0.55) · Type: tuple[float, float]

Target range for acceptance rate.

Writers (writers.*)#

No external writers are enabled by default. If you enable them manually, the root-level writer keys below control their configuration. The evaluation stage always writes per-step statistics to an internal HDF5 file for digest computation; this is independent of the writers configured here.

Console writer (writers.console.*)#

writers.console.interval

Default: 1 · Type: int

Step interval for logging.

writers.console.fields

Default: 'loss' · Type: str

Comma-separated list of field specs.

CSV writer (writers.csv.*)#

writers.csv.path_template

Default: '{stage}_stats.csv' · Type: str

Output path template.

HDF5 writer (writers.hdf5.*)#

writers.hdf5.path_template

Default: '{stage}_stats.h5' · Type: str

Output path template.

Estimators (estimators.*)#

Energy estimator definitions match training, with additional evaluation-only estimators enabled through boolean flags.

  • PotentialEnergy and TotalEnergy are added automatically by the workflow and are not configurable via config keys.

  • estimators.enabled.energy defaults to true.

  • estimators.enabled.spin defaults to false.

  • estimators.enabled.density defaults to false.

Kinetic energy (estimators.energy.kinetic.*)#

estimators.energy.kinetic.mode

Default: forward_laplacian · Type: LaplacianMode

Laplacian computation strategy. forward_laplacian is the default for JAX 0.7.1 and later, scan for earlier versions.

estimators.energy.kinetic.sparsity_threshold

Default: 0 · Type: int

Sparsity threshold when using forward_laplacian mode.

ECP energy (estimators.energy.ecp.*)#

Added automatically when system.ecp is set.

estimators.energy.ecp.max_core

Default: 2 · Type: int

Maximum number of nearest ECP atoms to consider per electron when evaluating nonlocal integrals.

estimators.energy.ecp.quadrature_id

Default: None · Type: str | None

Spherical quadrature rule used to evaluate nonlocal ECP integrals.

estimators.energy.ecp.electrons_field

Default: 'electrons' · Type: str

Name of electron position field in data.

estimators.energy.ecp.atoms_field

Default: 'atoms' · Type: str

Name of atom position field in data.

Density (estimators.density.*)#

Electron density in fractional (lattice) coordinates. Converts Cartesian positions to fractional coordinates via the inverse lattice matrix, then histograms within \([0, 1)\).

When enabled without overrides, the workflow wires three independent 1-D histograms along the a, b, and c lattice directions, each with 50 bins. To keep only specific axes, set the others to null. Each axis override accepts lattice_index and bins.

# Just enable with defaults (a, b, c histograms, 50 bins each):
estimators:
  enabled:
    density: true

# Keep only c-axis with finer bins:
estimators:
  enabled:
    density: true
  density:
    axes:
      a: null
      b: null
      c:
        lattice_index: 2
        bins: 100