Merge pull request #332 from apax-hub/erbs_fixes

full ensemble fix, feature model fix

apax/config/md_config.py

@@ -1,12 +1,12 @@
 import os
-
-# from types import UnionType
 from typing import Literal, Union
 
 import yaml
 from pydantic import BaseModel, Field, NonNegativeInt, PositiveFloat, PositiveInt
 from typing_extensions import Annotated
 
+from apax.utils.helpers import APAX_PROPERTIES
+
 
 class ConstantTempSchedule(BaseModel, extra="forbid"):
     """Constant temperature schedule.
@@ -234,6 +234,14 @@ class MDConfig(BaseModel, frozen=True, extra="forbid"):
     extra_capacity : int, default = 0
         | JaxMD allocates a maximal number of neighbors. This argument lets you add
         | additional capacity to avoid recompilation. The default is usually fine.
+
+    dynamics_checks: list[DynamicsCheck]
+        | List of termination criteria. Currently energy and force uncertainty
+        | are available
+    properties: list[str]
+        | Whitelist of properties to be saved in the trajectory.
+        | This does not effect what the model will calculate, e.g..
+        | an ensemble will still calculate uncertainties.
     initial_structure : str, required
         | Path to the starting structure of the simulation.
     sim_dir : str, default = "."
@@ -266,6 +274,8 @@ class MDConfig(BaseModel, frozen=True, extra="forbid"):
 
     dynamics_checks: list[DynamicsCheck] = []
 
+    properties: list[str] = APAX_PROPERTIES
+
     initial_structure: str
     load_momenta: bool = False
     sim_dir: str = "."

apax/md/ase_calc.py

@@ -82,8 +82,18 @@ def make_ensemble(model):
     def ensemble(positions, Z, idx, box, offsets):
         results = model(positions, Z, idx, box, offsets)
         uncertainty = {k + "_uncertainty": jnp.std(v, axis=0) for k, v in results.items()}
+        ensemble = {k + "_ensemble": v for k, v in results.items()}
         results = {k: jnp.mean(v, axis=0) for k, v in results.items()}
+        if "forces_ensemble" in ensemble.keys():
+            ensemble["forces_ensemble"] = jnp.transpose(
+                ensemble["forces_ensemble"], (1, 2, 0)
+            )
+        if "forces_ensemble" in ensemble.keys():
+            ensemble["stress_ensemble"] = jnp.transpose(
+                ensemble["forces_ensemble"], (1, 2, 0)
+            )
         results.update(uncertainty)
+        results.update(ensemble)
 
         return results
 

apax/md/io.py

@@ -8,18 +8,29 @@
 from ase.calculators.singlepoint import SinglePointCalculator
 
 from apax.md.sim_utils import System
+from apax.utils.helpers import APAX_PROPERTIES
 from apax.utils.jax_md_reduced import space
 
 log = logging.getLogger(__name__)
 
 
 class TrajHandler:
-    def __init__(self) -> None:
-        self.system: System
-        self.sampling_rate: int
-        self.buffer_size: int
-        self.traj_path: Path
-        self.time_step: float
+    def __init__(
+        self,
+        system: System,
+        sampling_rate: int,
+        buffer_size: int,
+        traj_path: Path,
+        time_step: float = 0.5,
+        properties: list[str] = APAX_PROPERTIES,
+    ) -> None:
+        self.atomic_numbers = system.atomic_numbers
+        self.box = system.box
+        self.fractional = np.any(self.box > 1e-6)
+        self.sampling_rate = sampling_rate
+        self.traj_path = traj_path
+        self.time_step = time_step
+        self.properties = properties
 
     def step(self, state_and_energy, transform=None):
         pass
@@ -53,6 +64,7 @@ def atoms_from_state(self, state, predictions, nbr_kwargs):
         atoms.pbc = np.diag(atoms.cell.array) > 1e-6
         predictions = {k: np.array(v) for k, v in predictions.items()}
         predictions["energy"] = predictions["energy"].item()
+        predictions = {k: v for k, v in predictions.items() if k in self.properties}
         atoms.calc = SinglePointCalculator(atoms, **predictions)
         return atoms
 
@@ -65,13 +77,15 @@ def __init__(
         buffer_size: int,
         traj_path: Path,
         time_step: float = 0.5,
+        properties: list[str] = [],
     ) -> None:
         self.atomic_numbers = system.atomic_numbers
         self.box = system.box
         self.fractional = np.any(self.box > 1e-6)
         self.sampling_rate = sampling_rate
         self.traj_path = traj_path
         self.time_step = time_step
+        self.properties = properties
         self.db = znh5md.IO(
             self.traj_path, timestep=self.time_step, store="time", save_units=False
         )

apax/md/simulate.py

@@ -166,10 +166,10 @@ def run_sim(
     n_inner: int,
     extra_capacity: int,
     rng_key: int,
+    traj_handler: TrajHandler,
     load_momenta: bool = False,
     restart: bool = True,
     checkpoint_interval: int = 50_000,
-    traj_handler: TrajHandler = TrajHandler(),
     dynamics_checks: list[DynamicsCheckBase] = [],
     disable_pbar: bool = False,
 ):
@@ -520,6 +520,7 @@ def run_md(model_config: Config, md_config: MDConfig, log_level="error"):
         md_config.buffer_size,
         traj_path,
         md_config.ensemble.dt,
+        properties=md_config.properties,
     )
     # TODO implement correct chunking
 
@@ -531,10 +532,10 @@ def run_md(model_config: Config, md_config: MDConfig, log_level="error"):
         n_inner=md_config.n_inner,
         extra_capacity=md_config.extra_capacity,
         load_momenta=md_config.load_momenta,
+        traj_handler=traj_handler,
         rng_key=jax.random.PRNGKey(md_config.seed),
         restart=md_config.restart,
         checkpoint_interval=md_config.checkpoint_interval,
         sim_dir=sim_dir,
-        traj_handler=traj_handler,
         dynamics_checks=dynamics_checks,
     )

apax/nn/models.py

@@ -79,8 +79,10 @@ def __call__(
             perturbation,
         )
 
-        gm = self.descriptor(dr_vec, Z, idx)
-        features = jax.vmap(self.readout)(gm)
+        features = self.descriptor(dr_vec, Z, idx)
+
+        if self.readout:
+            features = jax.vmap(self.readout)(features)
 
         if self.mask_atoms:
             features = mask_by_atom(features, Z)
@@ -268,7 +270,9 @@ def __call__(
 
             prediction["forces"] = forces_mean
             prediction["forces_uncertainty"] = jnp.sqrt(forces_variance)
-            prediction["forces_ensemble"] = forces_ens
+
+            forces_ens = jnp.transpose(forces_ens, (1, 2, 0))
+            prediction["forces_ensemble"] = forces_ens  # n_atoms x 3 x n_members
 
         else:
             forces_mean = -jax.grad(mean_energy_fn)(R, Z, neighbor, box, offsets)

apax/utils/helpers.py

@@ -2,22 +2,26 @@
 
 import yaml
 
+APAX_PROPERTIES = [
+    "energy",
+    "forces",
+    "stress",
+    "forces_uncertainty",
+    "energy_uncertainty",
+    "stress_uncertainty",
+    "energy_ensemble",
+    "forces_ensemble",
+    "stress_ensemble",
+]
+
 
 def setup_ase():
     """Add uncertainty keys to ASE all properties.
     from https://github.com/zincware/IPSuite/blob/main/ipsuite/utils/helpers.py#L10
     """
     from ase.calculators.calculator import all_properties
 
-    additional_keys = [
-        "forces_uncertainty",
-        "energy_uncertainty",
-        "stress_uncertainty",
-        "energy_ensemble",
-        "forces_ensemble",
-    ]
-
-    for val in additional_keys:
+    for val in APAX_PROPERTIES:
         if val not in all_properties:
             all_properties.append(val)
 

tests/integration_tests/md/md_config_threshold.yaml

@@ -1,17 +1,22 @@
 ensemble:
   name: nvt
-  dt: 0.1 # fs time step
+  dt: 0.2 # fs time step
   temperature_schedule:
     name: piecewise
-    T0: 5 # K
+    T0: 50 # K
     values: [100, 200, 1000]
     steps: [10, 20, 30]
 
-duration: 100 # fs
+duration: 500 # fs
 n_inner: 1
 sampling_rate: 1
 checkpoint_interval: 2
 restart: True
 dynamics_checks:
   - name: forces_uncertainty
-    threshold: 1.0
+    threshold: 0.01
+properties:
+  - energy
+  - forces
+  - energy_uncertainty
+  - forces_ensemble

tests/integration_tests/md/test_md.py

@@ -200,7 +200,7 @@ def test_jaxmd_schedule_and_thresold(get_tmp_path, example_dataset):
     }
     model_config_dict = load_config_and_run_training(model_confg_path, data_config_mods)
 
-    md_confg_path = TEST_PATH / "md_config.yaml"
+    md_confg_path = TEST_PATH / "md_config_threshold.yaml"
 
     with open(md_confg_path.as_posix(), "r") as stream:
         md_config_dict = yaml.safe_load(stream)
@@ -214,3 +214,13 @@ def test_jaxmd_schedule_and_thresold(get_tmp_path, example_dataset):
 
     traj = znh5md.IO(md_config.sim_dir + "/" + md_config.traj_name)[:]
     assert len(traj) < 1000  # num steps
+
+    results_keys = list(traj[0].calc.results.keys())
+
+    assert "energy" in results_keys
+    assert "forces" in results_keys
+    assert "energy_uncertainty" in results_keys
+    assert "forces_ensemble" in results_keys
+
+    assert "energy_ensemble" not in results_keys
+    assert "forces_uncertainty" not in results_keys