chore(tests): ensure the same result of frame 0 and 1 (deepmodeling#4442

)

Copied from njzjz/deepmd-gnn#27.

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

- **Bug Fixes**
- Enhanced the robustness of the testing framework to ensure consistent
model output across multiple frames of input data.
- Added assertions to validate output equivalence for the first and
second frames.

- **Tests**
- Adjusted the testing methods to accommodate changes in input
dimensionality and ensure proper validation of model behavior.
<!-- end of auto-generated comment: release notes by coderabbit.ai -->

---------

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

source/tests/universal/common/cases/model/utils.py

@@ -119,18 +119,18 @@ def test_has_message_passing(self) -> None:
     def test_forward(self) -> None:
         """Test forward and forward_lower."""
         test_spin = getattr(self, "test_spin", False)
-        nf = 1
+        nf = 2
         natoms = 5
         aprec = (
             0
             if self.aprec_dict.get("test_forward", None) is None
             else self.aprec_dict["test_forward"]
         )
         rng = np.random.default_rng(GLOBAL_SEED)
-        coord = 4.0 * rng.random([natoms, 3]).reshape([nf, -1])
-        atype = np.array([0, 0, 0, 1, 1], dtype=int).reshape([nf, -1])
-        spin = 0.5 * rng.random([natoms, 3]).reshape([nf, -1])
-        cell = 6.0 * np.eye(3).reshape([nf, 9])
+        coord = 4.0 * rng.random([1, natoms, 3]).repeat(nf, 0).reshape([nf, -1])
+        atype = np.array([[0, 0, 0, 1, 1] * nf], dtype=int).reshape([nf, -1])
+        spin = 0.5 * rng.random([1, natoms, 3]).repeat(nf, 0).reshape([nf, -1])
+        cell = 6.0 * np.repeat(np.eye(3)[None, ...], nf, axis=0).reshape([nf, 9])
         coord_ext, atype_ext, mapping, nlist = extend_input_and_build_neighbor_list(
             coord,
             atype,
@@ -147,9 +147,9 @@ def test_forward(self) -> None:
         aparam = None
         fparam = None
         if self.module.get_dim_aparam() > 0:
-            aparam = rng.random([nf, natoms, self.module.get_dim_aparam()])
+            aparam = rng.random([1, natoms, self.module.get_dim_aparam()]).repeat(nf, 0)
         if self.module.get_dim_fparam() > 0:
-            fparam = rng.random([nf, self.module.get_dim_fparam()])
+            fparam = rng.random([1, self.module.get_dim_fparam()]).repeat(nf, 0)
         ret = []
         ret_lower = []
         for module in self.modules_to_test:
@@ -183,6 +183,15 @@ def test_forward(self) -> None:
             ret_lower.append(module.forward_lower(**input_dict_lower))
 
         for kk in ret[0]:
+            # ensure the first frame and the second frame are the same
+            if ret[0][kk] is not None:
+                np.testing.assert_allclose(
+                    ret[0][kk][0],
+                    ret[0][kk][1],
+                    err_msg=f"compare {kk} between frame 0 and 1",
+                    atol=aprec,
+                )
+
             subret = []
             for rr in ret:
                 if rr is not None: