fix(finetune): calculate fitting stat when using random fitting in finetuning process

deepmd/pt/model/atomic_model/base_atomic_model.py

@@ -487,6 +487,23 @@ def change_out_bias(
         else:
             raise RuntimeError("Unknown bias_adjust_mode mode: " + bias_adjust_mode)
 
+    def compute_fitting_stat(self, sample_merged) -> None:
+        """Compute the input statistics (e.g. mean and stddev) for the fittings from packed data..
+
+        Parameters
+        ----------
+        sample_merged : Union[Callable[[], list[dict]], list[dict]]
+            - list[dict]: A list of data samples from various data systems.
+                Each element, `merged[i]`, is a data dictionary containing `keys`: `torch.Tensor`
+                originating from the `i`-th data system.
+            - Callable[[], list[dict]]: A lazy function that returns data samples in the above format
+                only when needed. Since the sampling process can be slow and memory-intensive,
+                the lazy function helps by only sampling once.
+        """
+        self.fitting_net.compute_input_stats(
+            sample_merged, protection=self.data_stat_protect
+        )
+
     def _get_forward_wrapper_func(self) -> Callable[..., torch.Tensor]:
         """Get a forward wrapper of the atomic model for output bias calculation."""
 

deepmd/pt/model/atomic_model/dp_atomic_model.py

@@ -324,9 +324,7 @@ def wrapped_sampler():
             return sampled
 
         self.descriptor.compute_input_stats(wrapped_sampler, stat_file_path)
-        self.fitting_net.compute_input_stats(
-            wrapped_sampler, protection=self.data_stat_protect
-        )
+        self.compute_fitting_stat(wrapped_sampler)
         if compute_or_load_out_stat:
             self.compute_or_load_out_stat(wrapped_sampler, stat_file_path)
 

deepmd/pt/model/model/make_model.py

@@ -230,6 +230,8 @@ def change_out_bias(
                 merged,
                 bias_adjust_mode=bias_adjust_mode,
             )
+            if bias_adjust_mode == "set-by-statistic":
+                self.atomic_model.compute_fitting_stat(merged)
 
         def forward_common_lower(
             self,