diff --git a/deepmd/pt/utils/dataloader.py b/deepmd/pt/utils/dataloader.py
index 0359071d71..361bc4b0b6 100644
--- a/deepmd/pt/utils/dataloader.py
+++ b/deepmd/pt/utils/dataloader.py
@@ -14,6 +14,7 @@
 )
 
 import h5py
+import numpy as np
 import torch
 import torch.distributed as dist
 import torch.multiprocessing
@@ -106,29 +107,34 @@ def construct_dataset(system):
 
         self.dataloaders = []
         self.batch_sizes = []
-        for system in self.systems:
+        if isinstance(batch_size, str):
+            if batch_size == "auto":
+                rule = 32
+            elif batch_size.startswith("auto:"):
+                rule = int(batch_size.split(":")[1])
+            else:
+                rule = None
+                log.error("Unsupported batch size type")
+            for ii in self.systems:
+                ni = ii._natoms
+                bsi = rule // ni
+                if bsi * ni < rule:
+                    bsi += 1
+                self.batch_sizes.append(bsi)
+        elif isinstance(batch_size, list):
+            self.batch_sizes = batch_size
+        else:
+            self.batch_sizes = batch_size * np.ones(len(systems), dtype=int)
+        assert len(self.systems) == len(self.batch_sizes)
+        for system, batch_size in zip(self.systems, self.batch_sizes):
             if dist.is_initialized():
                 system_sampler = DistributedSampler(system)
                 self.sampler_list.append(system_sampler)
             else:
                 system_sampler = None
-            if isinstance(batch_size, str):
-                if batch_size == "auto":
-                    rule = 32
-                elif batch_size.startswith("auto:"):
-                    rule = int(batch_size.split(":")[1])
-                else:
-                    rule = None
-                    log.error("Unsupported batch size type")
-                self.batch_size = rule // system._natoms
-                if self.batch_size * system._natoms < rule:
-                    self.batch_size += 1
-            else:
-                self.batch_size = batch_size
-            self.batch_sizes.append(self.batch_size)
             system_dataloader = DataLoader(
                 dataset=system,
-                batch_size=self.batch_size,
+                batch_size=int(batch_size),
                 num_workers=0,  # Should be 0 to avoid too many threads forked
                 sampler=system_sampler,
                 collate_fn=collate_batch,