support PBC

Signed-off-by: Jinzhe Zeng <[email protected]>

deepmd/pt/model/model/mace.py

@@ -435,17 +435,38 @@ def forward_lower(
             model_predict["extended_virial"] = model_ret["energy_derv_c"].squeeze(-3)
         return model_predict
 
-    def forward_lower_common(
+    def forward_lower_common(  # noqa: PLR0915
         self,
-        extended_coord,
-        extended_atype,
-        nlist,
-        mapping: Optional[torch.Tensor] = None,
+        extended_coord: torch.Tensor,
+        extended_atype: torch.Tensor,
+        nlist: torch.Tensor,
+        mapping: Optional[torch.Tensor] = None,  # noqa: ARG002
         fparam: Optional[torch.Tensor] = None,
         aparam: Optional[torch.Tensor] = None,
-        do_atomic_virial: bool = False,
-        comm_dict: Optional[Dict[str, torch.Tensor]] = None,
-    ):
+        do_atomic_virial: bool = False,  # noqa: ARG002
+        comm_dict: Optional[dict[str, torch.Tensor]] = None,
+    ) -> dict[str, torch.Tensor]:
+        """Forward lower common pass of the model.
+
+        Parameters
+        ----------
+        extended_coord : torch.Tensor
+            The extended coordinates of atoms.
+        extended_atype : torch.Tensor
+            The extended atomic types of atoms.
+        nlist : torch.Tensor
+            The neighbor list.
+        mapping : torch.Tensor, optional
+            The mapping tensor.
+        fparam : torch.Tensor, optional
+            The frame parameters.
+        aparam : torch.Tensor, optional
+            The atomic parameters.
+        do_atomic_virial : bool, optional
+            Whether to compute atomic virial.
+        comm_dict : dict[str, torch.Tensor], optional
+            The communication dictionary.
+        """
         extended_coord_ = extended_coord
         nf, nall, _ = extended_coord_.shape
         _, nloc, _ = nlist.shape
@@ -515,21 +536,35 @@ def forward_lower_common(
                         device=extended_coord_ff.device,
                     ),
                 },
-                compute_virials=True,
+                compute_force=False,
+                compute_virials=False,
+                compute_stress=False,
+                compute_displacement=True,
                 training=self.training,
             )
-            energy = ret["energy"]
-            assert energy is not None
-            energy = energy.view(1, 1).to(extended_coord_.dtype)
-            force = ret["forces"]
-            assert force is not None
-            force = force.view(1, nall, 3).to(extended_coord_.dtype)
-            virial = ret["virials"]
-            assert virial is not None
-            virial = virial.view(1, 9)
+
             atom_energy = ret["node_energy"]
-            assert atom_energy is not None
-            atom_energy = atom_energy.view(1, nall).to(extended_coord_.dtype)[:, :nall]
+            if atom_energy is None:
+                msg = "atom_energy is None"
+                raise ValueError(msg)
+            atom_energy = atom_energy.view(1, nall).to(extended_coord_.dtype)[:, :nloc]
+            energy = torch.sum(atom_energy, dim=1).view(1, 1).to(extended_coord_.dtype)
+            grad_outputs: list[Optional[torch.Tensor]] = [
+                torch.ones_like(energy),
+            ]
+            displacement = ret["displacement"]
+            force, virial = torch.autograd.grad(
+                outputs=[energy],
+                inputs=[extended_coord_ff, displacement],
+                grad_outputs=grad_outputs,
+                retain_graph=True,
+                create_graph=self.training,
+            )
+            force = -force
+            virial = -virial
+            force = force.view(1, nall, 3).to(extended_coord_.dtype)
+            virial = virial.view(1, 9).to(extended_coord_.dtype)
+
             energies.append(energy)
             forces.append(force)
             virials.append(virial)
@@ -544,7 +579,7 @@ def forward_lower_common(
             "energy_derv_r": forces.view(nf, nall, 1, 3),
             "energy_derv_c_redu": virials.view(nf, 1, 9),
             # take the first nloc atoms to match other models
-            "energy": atom_energies.view(nf, nall, 1)[:, :nloc, :],
+            "energy": atom_energies.view(nf, nloc, 1),
             # fake atom_virial
             "energy_derv_c": torch.zeros(
                 (nf, nall, 1, 9),

examples/water/mace/input.json

@@ -0,0 +1,60 @@
+{
+  "_comment1": " model parameters",
+  "model": {
+    "type": "mace",
+    "type_map": [
+      "O",
+      "H"
+    ],
+    "r_max": 6.0,
+    "sel": "auto",
+    "_comment2": " that's all"
+  },
+
+  "learning_rate": {
+    "type": "exp",
+    "decay_steps": 5000,
+    "start_lr": 0.001,
+    "stop_lr": 3.51e-8,
+    "_comment5": "that's all"
+  },
+
+  "loss": {
+    "type": "ener",
+    "start_pref_e": 0.02,
+    "limit_pref_e": 1,
+    "start_pref_f": 1000,
+    "limit_pref_f": 1,
+    "start_pref_v": 0,
+    "limit_pref_v": 0,
+    "_comment6": " that's all"
+  },
+
+  "training": {
+    "training_data": {
+      "systems": [
+        "../data/data_0/",
+        "../data/data_1/",
+        "../data/data_2/"
+      ],
+      "batch_size": "auto",
+      "_comment7": "that's all"
+    },
+    "validation_data": {
+      "systems": [
+        "../data/data_3"
+      ],
+      "batch_size": 1,
+      "numb_btch": 3,
+      "_comment8": "that's all"
+    },
+    "numb_steps": 1000000,
+    "seed": 10,
+    "disp_file": "lcurve.out",
+    "disp_freq": 100,
+    "save_freq": 1000,
+    "_comment9": "that's all"
+  },
+
+  "_comment10": "that's all"
+}