Source code for torch_sim.runners

"""High level runners for atomistic simulations.

This module provides functions for running molecular dynamics simulations and geometry
optimizations using various models and integrators. It includes utilities for
converting between different atomistic representations and handling simulation state.
"""

import warnings
from collections.abc import Callable
from dataclasses import dataclass
from itertools import chain

import torch

from torch_sim.autobatching import BinningAutoBatcher, InFlightAutoBatcher
from torch_sim.models.interface import ModelInterface
from torch_sim.quantities import batchwise_max_force, calc_kinetic_energy, calc_kT
from torch_sim.state import SimState, concatenate_states, initialize_state
from torch_sim.trajectory import TrajectoryReporter
from torch_sim.typing import StateLike
from torch_sim.units import UnitSystem


def _configure_reporter(
    trajectory_reporter: TrajectoryReporter | dict | None,
    *,
    state_kwargs: dict | None = None,
    properties: list[str] | None = None,
    prop_frequency: int = 10,
    state_frequency: int = 100,
) -> TrajectoryReporter:
    if trajectory_reporter is None:
        return None
    if isinstance(trajectory_reporter, TrajectoryReporter):
        return trajectory_reporter
    possible_properties = {
        "potential_energy": lambda state: state.energy,
        "forces": lambda state: state.forces,
        "stress": lambda state: state.stress,
        "kinetic_energy": lambda state: calc_kinetic_energy(state.momenta, state.masses),
        "temperature": lambda state: calc_kT(state.momenta, state.masses),
    }

    prop_calculators = {
        prop: calculator
        for prop, calculator in possible_properties.items()
        if prop in properties
    }

    # ordering is important to ensure we can override defaults
    return TrajectoryReporter(
        prop_calculators=trajectory_reporter.pop(
            "prop_calculators", {prop_frequency: prop_calculators}
        ),
        state_frequency=trajectory_reporter.pop("state_frequency", state_frequency),
        state_kwargs=state_kwargs or {},
        **trajectory_reporter,
    )


def _configure_batches_iterator(
    model: ModelInterface,
    state: SimState,
    autobatcher: BinningAutoBatcher | bool,
) -> BinningAutoBatcher | list[tuple[SimState, list[int]]]:
    """Create a batches iterator for the integrate function.

    Args:
        model (ModelInterface): The model to use for the integration
        state (SimState): The state to use for the integration
        autobatcher (BinningAutoBatcher | bool): The autobatcher to use for integration

    Returns:
        A batches iterator
    """
    # load and properly configure the autobatcher
    if autobatcher is True:
        autobatcher = BinningAutoBatcher(
            model=model,
            return_indices=True,
            max_memory_padding=0.9,
        )
        autobatcher.load_states(state)
        batches = autobatcher
    elif isinstance(autobatcher, BinningAutoBatcher):
        autobatcher.load_states(state)
        autobatcher.return_indices = True
        batches = autobatcher
    elif autobatcher is False:
        batches = [(state, [])]
    else:
        raise TypeError(
            f"Invalid autobatcher type: {type(autobatcher).__name__}, "
            "must be bool or BinningAutoBatcher."
        )
    return batches


[docs] def integrate( system: StateLike, model: ModelInterface, *, integrator: Callable, n_steps: int, temperature: float | list | torch.Tensor, timestep: float, trajectory_reporter: TrajectoryReporter | dict | None = None, autobatcher: BinningAutoBatcher | bool = False, **integrator_kwargs: dict, ) -> SimState: """Simulate a system using a model and integrator. Args: system (StateLike): Input system to simulate model (ModelInterface): Neural network model module integrator (Callable): Integration algorithm function n_steps (int): Number of integration steps temperature (float | ArrayLike): Temperature or array of temperatures for each step timestep (float): Integration time step integrator_kwargs: Additional keyword arguments for integrator trajectory_reporter (TrajectoryReporter | dict | None): Optional reporter for tracking trajectory. If a dict, will be passed to the TrajectoryReporter constructor. autobatcher (BinningAutoBatcher | bool): Optional autobatcher to use **integrator_kwargs: Additional keyword arguments for integrator init function Returns: SimState: Final state after integration """ unit_system = UnitSystem.metal # create a list of temperatures temps = temperature if hasattr(temperature, "__iter__") else [temperature] * n_steps if len(temps) != n_steps: raise ValueError(f"{len(temps)=:,}. It must equal n_steps = {n_steps=:,}") # initialize the state state: SimState = initialize_state(system, model.device, model.dtype) dtype, device = state.dtype, state.device kTs = torch.tensor(temps, dtype=dtype, device=device) * unit_system.temperature init_fn, update_fn = integrator( model=model, kT=kTs[0], dt=torch.tensor(timestep * unit_system.time, dtype=dtype, device=device), **integrator_kwargs, ) # batch_iterator will be a list if autobatcher is False batch_iterator = _configure_batches_iterator(model, state, autobatcher) trajectory_reporter = _configure_reporter( trajectory_reporter, properties=["kinetic_energy", "potential_energy", "temperature"], ) final_states: list[SimState] = [] og_filenames = trajectory_reporter.filenames if trajectory_reporter else None for state, batch_indices in batch_iterator: state = init_fn(state) # set up trajectory reporters if autobatcher and trajectory_reporter: # we must remake the trajectory reporter for each batch trajectory_reporter.load_new_trajectories( filenames=[og_filenames[i] for i in batch_indices] ) # run the simulation for step in range(1, n_steps + 1): state = update_fn(state, kT=kTs[step - 1]) if trajectory_reporter: trajectory_reporter.report(state, step, model=model) # finish the trajectory reporter final_states.append(state) if trajectory_reporter: trajectory_reporter.finish() if isinstance(batch_iterator, BinningAutoBatcher): reordered_states = batch_iterator.restore_original_order(final_states) return concatenate_states(reordered_states) return state
def _configure_in_flight_autobatcher( model: ModelInterface, state: SimState, autobatcher: InFlightAutoBatcher | bool, max_attempts: int, # TODO: change name to max_iterations ) -> InFlightAutoBatcher: """Configure the hot swapping autobatcher for the optimize function. Args: model (ModelInterface): The model to use for the autobatcher state (SimState): The state to use for the autobatcher autobatcher (InFlightAutoBatcher | bool): The autobatcher to use for the autobatcher max_attempts (int): The maximum number of attempts for the autobatcher Returns: A hot swapping autobatcher """ # load and properly configure the autobatcher if isinstance(autobatcher, InFlightAutoBatcher): autobatcher.return_indices = True autobatcher.max_attempts = max_attempts elif isinstance(autobatcher, bool): if autobatcher: memory_scales_with = model.memory_scales_with max_memory_scaler = None else: memory_scales_with = "n_atoms" max_memory_scaler = state.n_atoms + 1 autobatcher = InFlightAutoBatcher( model=model, return_indices=True, max_memory_scaler=max_memory_scaler, memory_scales_with=memory_scales_with, max_iterations=max_attempts, max_memory_padding=0.9, ) else: autobatcher_type = type(autobatcher).__name__ cls_name = InFlightAutoBatcher.__name__ raise TypeError(f"Invalid {autobatcher_type=}, must be bool or {cls_name}.") return autobatcher def _chunked_apply( fn: Callable, states: SimState, model: ModelInterface, **batcher_kwargs: dict, ) -> SimState: """Apply a function to a state in chunks. This prevents us from running out of memory when applying a function to a large number of states. Args: fn (Callable): The function to apply states (SimState): The state to apply the function to model (ModelInterface): The model to use for the autobatcher **batcher_kwargs: Additional keyword arguments for the autobatcher Returns: A state with the function applied """ autobatcher = BinningAutoBatcher( model=model, return_indices=False, **batcher_kwargs, ) autobatcher.load_states(states) initialized_states = [] initialized_states = [fn(batch) for batch in autobatcher] ordered_states = autobatcher.restore_original_order(initialized_states) return concatenate_states(ordered_states)
[docs] def generate_force_convergence_fn(force_tol: float = 1e-1) -> Callable: """Generate a force-based convergence function for the convergence_fn argument of the optimize function. Args: force_tol (float): Force tolerance for convergence Returns: Convergence function that takes a state and last energy and returns a batchwise boolean function """ def convergence_fn( state: SimState, last_energy: torch.Tensor | None = None, # noqa: ARG001 ) -> bool: """Check if the system has converged.""" return batchwise_max_force(state) < force_tol return convergence_fn
[docs] def generate_energy_convergence_fn(energy_tol: float = 1e-3) -> Callable: """Generate an energy-based convergence function for the convergence_fn argument of the optimize function. Args: energy_tol (float): Energy tolerance for convergence Returns: Convergence function that takes a state and last energy and returns a batchwise boolean function """ def convergence_fn( state: SimState, last_energy: torch.Tensor | None = None, ) -> bool: """Check if the system has converged.""" return torch.abs(state.energy - last_energy) < energy_tol return convergence_fn
[docs] def optimize( system: StateLike, model: ModelInterface, *, optimizer: Callable, convergence_fn: Callable | None = None, trajectory_reporter: TrajectoryReporter | dict | None = None, autobatcher: InFlightAutoBatcher | bool = False, max_steps: int = 10_000, steps_between_swaps: int = 5, **optimizer_kwargs: dict, ) -> SimState: """Optimize a system using a model and optimizer. Args: system (StateLike): Input system to optimize (ASE Atoms, Pymatgen Structure, or SimState) model (ModelInterface): Neural network model module optimizer (Callable): Optimization algorithm function convergence_fn (Callable | None): Condition for convergence, should return a boolean tensor of length n_batches optimizer_kwargs: Additional keyword arguments for optimizer init function trajectory_reporter (TrajectoryReporter | dict | None): Optional reporter for tracking optimization trajectory. If a dict, will be passed to the TrajectoryReporter constructor. autobatcher (InFlightAutoBatcher | bool): Optional autobatcher to use. If False, the system will assume infinite memory and will not batch, but will still remove converged structures from the batch. If True, the system will estimate the memory available and batch accordingly. If a InFlightAutoBatcher, the system will use the provided autobatcher, but will reset the max_attempts to max_steps // steps_between_swaps. max_steps (int): Maximum number of total optimization steps steps_between_swaps: Number of steps to take before checking convergence and swapping out states. Returns: Optimized system state """ # create a default convergence function if one is not provided # TODO: document this behavior if convergence_fn is None: convergence_fn = generate_energy_convergence_fn(energy_tol=1e-3) # initialize the state state: SimState = initialize_state(system, model.device, model.dtype) init_fn, update_fn = optimizer(model=model, **optimizer_kwargs) max_attempts = max_steps // steps_between_swaps autobatcher = _configure_in_flight_autobatcher( model, state, autobatcher, max_attempts ) state = _chunked_apply( init_fn, state, model, max_memory_scaler=autobatcher.max_memory_scaler, memory_scales_with=autobatcher.memory_scales_with, ) autobatcher.load_states(state) trajectory_reporter = _configure_reporter( trajectory_reporter, properties=["potential_energy"], ) step: int = 1 last_energy = None all_converged_states, convergence_tensor = [], None og_filenames = trajectory_reporter.filenames if trajectory_reporter else None while (result := autobatcher.next_batch(state, convergence_tensor))[0] is not None: state, converged_states, batch_indices = result all_converged_states.extend(converged_states) # need to update the trajectory reporter if any states have converged if trajectory_reporter and (step == 1 or len(converged_states) > 0): trajectory_reporter.load_new_trajectories( filenames=[og_filenames[i] for i in batch_indices] ) for _step in range(steps_between_swaps): last_energy = state.energy state = update_fn(state) if trajectory_reporter: trajectory_reporter.report(state, step, model=model) step += 1 if step > max_steps: # TODO: max steps should be tracked for each structure in the batch warnings.warn(f"Optimize has reached max steps: {step}", stacklevel=2) break convergence_tensor = convergence_fn(state, last_energy) all_converged_states.extend(result[1]) if trajectory_reporter: trajectory_reporter.finish() if autobatcher: final_states = autobatcher.restore_original_order(all_converged_states) return concatenate_states(final_states) return state
[docs] def static( system: StateLike, model: ModelInterface, *, trajectory_reporter: TrajectoryReporter | dict | None = None, autobatcher: BinningAutoBatcher | bool = False, ) -> list[dict[str, torch.Tensor]]: """Run single point calculations on a batch of systems. Unlike the other runners, this function does not return a state. Instead, it returns a list of dictionaries, one for each batch in the input state. Each dictionary contains the properties calculated for that batch. It will also modify the state in place with the "energy", "forces", and "stress" properties if they are present in the model output. Args: system (StateLike): Input system to calculate properties for model (ModelInterface): Neural network model module unit_system (UnitSystem): Unit system for energy and forces trajectory_reporter (TrajectoryReporter | dict | None): Optional reporter for tracking trajectory. If a dict, will be passed to the TrajectoryReporter constructor and must include at least the "filenames" key. Any prop calculators will be executed and the results will be returned in a list. Make sure that if multiple unique states are used, that the `variable_atomic_numbers` and `variable_masses` are set to `True` in the `state_kwargs` argument. autobatcher (BinningAutoBatcher | bool): Optional autobatcher to use for batching calculations Returns: list[dict[str, torch.Tensor]]: Maps of property names to tensors for all batches """ # initialize the state state: SimState = initialize_state(system, model.device, model.dtype) batch_iterator = _configure_batches_iterator(model, state, autobatcher) properties = ["potential_energy"] if model.compute_forces: properties.append("forces") if model.compute_stress: properties.append("stress") trajectory_reporter = _configure_reporter( trajectory_reporter or dict(filenames=None), state_kwargs={ "variable_atomic_numbers": True, "variable_masses": True, "save_forces": model.compute_forces, }, properties=properties, ) @dataclass class StaticState(type(state)): energy: torch.Tensor forces: torch.Tensor | None stress: torch.Tensor | None final_states: list[SimState] = [] all_props: list[dict[str, torch.Tensor]] = [] og_filenames = trajectory_reporter.filenames for substate, batch_indices in batch_iterator: # set up trajectory reporters if autobatcher and trajectory_reporter and og_filenames is not None: # we must remake the trajectory reporter for each batch trajectory_reporter.load_new_trajectories( filenames=[og_filenames[idx] for idx in batch_indices] ) model_outputs = model(substate) substate = StaticState( **vars(substate), energy=model_outputs["energy"], forces=model_outputs["forces"] if model.compute_forces else None, stress=model_outputs["stress"] if model.compute_stress else None, ) props = trajectory_reporter.report(substate, 0, model=model) all_props.extend(props) final_states.append(substate) trajectory_reporter.finish() if isinstance(batch_iterator, BinningAutoBatcher): # reorder properties to match original order of states original_indices = list(chain.from_iterable(batch_iterator.index_bins)) return [all_props[idx] for idx in original_indices] return all_props