Fix PaddlePaddle tensor construction warnings and enhance PyTorch to_torch_tensor function

Co-authored-by: njzjz <9496702+njzjz@users.noreply.github.com>

Author: Copilot

deepmd/pd/utils/utils.py

@@ -267,6 +267,9 @@ def to_paddle_tensor(
     if xx is None:
         return None
     assert xx is not None
+    # Handle PaddlePaddle tensors - clone and move to target device/dtype if needed
+    if isinstance(xx, paddle.Tensor):
+        return xx.clone().detach().to(device=DEVICE)
     if not isinstance(xx, np.ndarray):
         return xx
     # Create a reverse mapping of NP_PRECISION_DICT

source/tests/pd/common.py

@@ -79,7 +79,10 @@ def eval_model(
         if spins is not None:
             assert isinstance(spins, paddle.Tensor), err_msg
         assert isinstance(atom_types, paddle.Tensor) or isinstance(atom_types, list)
-        atom_types = paddle.to_tensor(atom_types, dtype=paddle.int32, place=DEVICE)
+        if isinstance(atom_types, paddle.Tensor):
+            atom_types = atom_types.clone().detach().to(dtype=paddle.int32, place=DEVICE)
+        else:
+            atom_types = paddle.to_tensor(atom_types, dtype=paddle.int32, place=DEVICE)
     elif isinstance(coords, np.ndarray):
         if cells is not None:
             assert isinstance(cells, np.ndarray), err_msg
@@ -101,28 +104,40 @@ def eval_model(
     else:
         natoms = len(atom_types[0])
 
-    coord_input = paddle.to_tensor(
-        coords.reshape([-1, natoms, 3]), dtype=GLOBAL_PD_FLOAT_PRECISION, place=DEVICE
-    )
+    if isinstance(coords, paddle.Tensor):
+        coord_input = coords.reshape([-1, natoms, 3]).clone().detach().to(dtype=GLOBAL_PD_FLOAT_PRECISION, place=DEVICE)
+    else:
+        coord_input = paddle.to_tensor(
+            coords.reshape([-1, natoms, 3]), dtype=GLOBAL_PD_FLOAT_PRECISION, place=DEVICE
+        )
     spin_input = None
     if spins is not None:
-        spin_input = paddle.to_tensor(
-            spins.reshape([-1, natoms, 3]),
-            dtype=GLOBAL_PD_FLOAT_PRECISION,
-            place=DEVICE,
-        )
+        if isinstance(spins, paddle.Tensor):
+            spin_input = spins.reshape([-1, natoms, 3]).clone().detach().to(dtype=GLOBAL_PD_FLOAT_PRECISION, place=DEVICE)
+        else:
+            spin_input = paddle.to_tensor(
+                spins.reshape([-1, natoms, 3]),
+                dtype=GLOBAL_PD_FLOAT_PRECISION,
+                place=DEVICE,
+            )
     has_spin = getattr(model, "has_spin", False)
     if callable(has_spin):
         has_spin = has_spin()
-    type_input = paddle.to_tensor(atom_types, dtype=paddle.int64, place=DEVICE)
+    if isinstance(atom_types, paddle.Tensor):
+        type_input = atom_types.clone().detach().to(dtype=paddle.int64, place=DEVICE)
+    else:
+        type_input = paddle.to_tensor(atom_types, dtype=paddle.int64, place=DEVICE)
     box_input = None
     if cells is None:
         pbc = False
     else:
         pbc = True
-        box_input = paddle.to_tensor(
-            cells.reshape([-1, 3, 3]), dtype=GLOBAL_PD_FLOAT_PRECISION, place=DEVICE
-        )
+        if isinstance(cells, paddle.Tensor):
+            box_input = cells.reshape([-1, 3, 3]).clone().detach().to(dtype=GLOBAL_PD_FLOAT_PRECISION, place=DEVICE)
+        else:
+            box_input = paddle.to_tensor(
+                cells.reshape([-1, 3, 3]), dtype=GLOBAL_PD_FLOAT_PRECISION, place=DEVICE
+            )
     num_iter = int((nframes + infer_batch_size - 1) / infer_batch_size)
 
     for ii in range(num_iter):