laser.measles.abm.snapshot

`laser.measles.abm.snapshot`

Snapshot save/load for the laser-measles ABM.

Snapshots capture the full population and patch state at a given point in time and allow the simulation to be resumed exactly from that point.

Typical usage::

import laser.measles as lm

# --- Segment 1 ---
model1 = lm.ABMModel(scenario, params1)
model1.components = [lm.VitalDynamicsProcess, lm.InfectionProcess]
model1.run()
lm.save_snapshot(model1, "checkpoint.h5")

# --- Segment 2 ---
params2 = lm.ABMParams(start_time="2001-01", num_ticks=365)
model2 = lm.load_snapshot("checkpoint.h5", params2,
                           components=[lm.VitalDynamicsProcess, lm.InfectionProcess])
model2.run()

Notes

save_snapshot modifies the model in place: R agents are squashed and vaccination dates are normalized. Do not continue using the model after saving.
The caller is responsible for setting params.start_time in the resumed segment to match the snapshot date (printed by load_snapshot).
WPPVitalDynamicsProcess is supported with the same guard mechanism.

`laser.measles.abm.snapshot.load_snapshot(path, params, components=None, verbose=True)`

Load an ABM from an HDF5 snapshot file and return it ready to run.

Restores the population, patch state, scenario, and metadata saved by save_snapshot. Components that modify the people frame (e.g. VitalDynamicsProcess) detect the snapshot context via model._from_snapshot and skip frame setup. Set params.start_time to the snapshot date printed by save_snapshot.

Parameters:

Name	Type	Description	Default
`path`	`str \| Path`	Path to the HDF5 snapshot file written by `save_snapshot`.	required
`params`	`ABMParams`	`ABMParams` for the resumed segment. Set `start_time` to the snapshot date and `num_ticks` to the remaining duration.	required
`components`	`list \| None`	Ordered list of component classes to attach — same list as used when building the original model.	`None`
`verbose`	`bool`	Print a loading summary.	`True`

Returns:

Type	Description
`ABMModel`	A configured `ABMModel` instance. Call `model.run()` to continue the simulation.

Example:

```python
import laser.measles as lm

params2 = lm.ABMParams(num_ticks=1825, seed=42, start_time="2009-12")
model2 = lm.load_snapshot(
    "checkpoint.h5",
    params2,
    components=[lm.VitalDynamicsProcess, lm.InfectionProcess],
)
model2.run()
```

`laser.measles.abm.snapshot.save_snapshot(model, path, squash_recovered=True, verbose=True)`

Save ABM state to an HDF5 snapshot file.

Call this after ABMModel.run() to persist the full population and patch state. The resulting HDF5 file can be resumed with load_snapshot to continue the simulation from exactly where it left off — useful for warm-start parameter sweeps, segmented cluster jobs, or reproducible checkpoints.

Warning

This function mutates model: recovered agents are squashed (if squash_recovered=True) and future vaccination dates are normalized. Do not continue running the model after calling this function.

Parameters:

Name	Type	Description	Default
`model`	`ABMModel`	A fully-run (or mid-run) `ABMModel` instance.	required
`path`	`str \| Path`	Destination HDF5 file path (created or overwritten).	required
`squash_recovered`	`bool`	If `True` (default), remove recovered agents before saving. This dramatically reduces file size for long measles runs.	`True`
`verbose`	`bool`	Print a progress summary.	`True`

Example:

```python
import laser.measles as lm

params = lm.ABMParams(num_ticks=3650, seed=42, start_time="2000-01")
model = lm.ABMModel(scenario, params)
model.components = [lm.VitalDynamicsProcess, lm.InfectionProcess]
model.run()

lm.save_snapshot(model, "checkpoint.h5")
# Do not use model after this point.
```