feat(pt): add Mean absolute percentage error (MAPE) loss for prop. pred. (#4854)

feat: add Mean absolute percentage error (MAPE) loss for property
prediction

<!-- This is an auto-generated comment: release notes by coderabbit.ai
-->
## Summary by CodeRabbit

* **New Features**
* Added support for the Mean Absolute Percentage Error (MAPE) as a
selectable loss function and metric.

* **Documentation**
* Updated user-facing documentation to include "mape" as an option for
loss function selection.
<!-- end of auto-generated comment: release notes by coderabbit.ai -->

---------

Co-authored-by: pre-commit-ci[bot] <66853113+pre-commit-ci[bot]@users.noreply.github.com>

Author: SchrodingersCattt

deepmd/pt/loss/property.py

@@ -42,7 +42,7 @@ def __init__(
         var_name : str
             The atomic property to fit, 'energy', 'dipole', and 'polar'.
         loss_func : str
-            The loss function, such as "smooth_mae", "mae", "rmse".
+            The loss function, such as "smooth_mae", "mae", "rmse", "mape".
         metric : list
             The metric such as mae, rmse which will be printed.
         beta : float
@@ -151,6 +151,12 @@ def forward(self, input_dict, model, label, natoms, learning_rate=0.0, mae=False
                     reduction="mean",
                 )
             )
+        elif self.loss_func == "mape":
+            loss += torch.mean(
+                torch.abs(
+                    (label[var_name] - model_pred[var_name]) / (label[var_name] + 1e-3)
+                )
+            )
         else:
             raise RuntimeError(f"Unknown loss function : {self.loss_func}")
 
@@ -182,6 +188,12 @@ def forward(self, input_dict, model, label, natoms, learning_rate=0.0, mae=False
                     reduction="mean",
                 )
             ).detach()
+        if "mape" in self.metric:
+            more_loss["mape"] = torch.mean(
+                torch.abs(
+                    (label[var_name] - model_pred[var_name]) / (label[var_name] + 1e-3)
+                )
+            ).detach()
 
         return model_pred, loss, more_loss