fix: fix for __array__ removal from array-api-strict

deepmd/dpmodel/common.py

@@ -8,6 +8,7 @@
     Optional,
 )
 
+import array_api_compat
 import ml_dtypes
 import numpy as np
 
@@ -105,7 +106,14 @@ def to_numpy_array(x: Any) -> Optional[np.ndarray]:
     """
     if x is None:
         return None
-    return np.asarray(x)
+    try:
+        # asarray is not within Array API standard, so may fail
+        return np.asarray(x)
+    except (ValueError, AttributeError):
+        xp = array_api_compat.array_namespace(x)
+        # to fix BufferError: Cannot export readonly array since signalling readonly is unsupported by DLPack.
+        x = xp.asarray(x, copy=True)
+        return np.from_dlpack(x)
 
 
 __all__ = [

deepmd/dpmodel/descriptor/dpa1.py

@@ -18,6 +18,9 @@
 from deepmd.dpmodel.array_api import (
     xp_take_along_axis,
 )
+from deepmd.dpmodel.common import (
+    to_numpy_array,
+)
 from deepmd.dpmodel.utils import (
     EmbeddingNet,
     EnvMat,
@@ -548,8 +551,8 @@ def serialize(self) -> dict:
             "exclude_types": obj.exclude_types,
             "env_protection": obj.env_protection,
             "@variables": {
-                "davg": np.array(obj["davg"]),
-                "dstd": np.array(obj["dstd"]),
+                "davg": to_numpy_array(obj["davg"]),
+                "dstd": to_numpy_array(obj["dstd"]),
             },
             ## to be updated when the options are supported.
             "trainable": self.trainable,
@@ -1022,8 +1025,8 @@ def serialize(self) -> dict:
             "exclude_types": obj.exclude_types,
             "env_protection": obj.env_protection,
             "@variables": {
-                "davg": np.array(obj["davg"]),
-                "dstd": np.array(obj["dstd"]),
+                "davg": to_numpy_array(obj["davg"]),
+                "dstd": to_numpy_array(obj["dstd"]),
             },
         }
         if obj.tebd_input_mode in ["strip"]:

pyproject.toml

@@ -85,8 +85,7 @@ test = [
     "pytest-sugar",
     "pytest-split",
     "dpgui",
-    # https://github.com/data-apis/array-api-strict/issues/85
-    'array-api-strict>=2,<2.1.1;python_version>="3.9"',
+    'array-api-strict>=2,!=2.1.1;python_version>="3.9"',
 ]
 docs = [
     "sphinx>=3.1.1",

source/tests/common/dpmodel/array_api/test_env_mat.py

@@ -14,7 +14,6 @@
 
 class TestEnvMat(unittest.TestCase, ArrayAPITest):
     def test_compute_smooth_weight(self):
-        self.set_array_api_version(compute_smooth_weight)
         d = xp.arange(10, dtype=xp.float64)
         w = compute_smooth_weight(
             d,

source/tests/consistent/descriptor/common.py

@@ -8,6 +8,9 @@
 from deepmd.common import (
     make_default_mesh,
 )
+from deepmd.dpmodel.common import (
+    to_numpy_array,
+)
 from deepmd.dpmodel.utils.nlist import (
     build_neighbor_list,
     extend_coord_with_ghosts,
@@ -141,7 +144,6 @@ def eval_array_api_strict_descriptor(
         box,
         mixed_types: bool = False,
     ) -> Any:
-        array_api_strict.set_array_api_strict_flags(api_version="2023.12")
         ext_coords, ext_atype, mapping = extend_coord_with_ghosts(
             array_api_strict.asarray(coords.reshape(1, -1, 3)),
             array_api_strict.asarray(atype.reshape(1, -1)),
@@ -157,7 +159,7 @@ def eval_array_api_strict_descriptor(
             distinguish_types=(not mixed_types),
         )
         return [
-            np.asarray(x) if hasattr(x, "__array_namespace__") else x
+            to_numpy_array(x) if hasattr(x, "__array_namespace__") else x
             for x in array_api_strict_obj(
                 ext_coords, ext_atype, nlist=nlist, mapping=mapping
             )

source/tests/consistent/fitting/test_dipole.py

@@ -6,6 +6,9 @@
 
 import numpy as np
 
+from deepmd.dpmodel.common import (
+    to_numpy_array,
+)
 from deepmd.dpmodel.fitting.dipole_fitting import DipoleFitting as DipoleFittingDP
 from deepmd.env import (
     GLOBAL_NP_FLOAT_PRECISION,
@@ -175,7 +178,7 @@ def eval_jax(self, jax_obj: Any) -> Any:
         )
 
     def eval_array_api_strict(self, array_api_strict_obj: Any) -> Any:
-        return np.asarray(
+        return to_numpy_array(
             array_api_strict_obj(
                 array_api_strict.asarray(self.inputs),
                 array_api_strict.asarray(self.atype.reshape(1, -1)),

source/tests/consistent/fitting/test_dos.py

@@ -6,6 +6,9 @@
 
 import numpy as np
 
+from deepmd.dpmodel.common import (
+    to_numpy_array,
+)
 from deepmd.dpmodel.fitting.dos_fitting import DOSFittingNet as DOSFittingDP
 from deepmd.env import (
     GLOBAL_NP_FLOAT_PRECISION,
@@ -218,7 +221,6 @@ def eval_jax(self, jax_obj: Any) -> Any:
         )
 
     def eval_array_api_strict(self, array_api_strict_obj: Any) -> Any:
-        array_api_strict.set_array_api_strict_flags(api_version="2023.12")
         (
             resnet_dt,
             precision,
@@ -227,7 +229,7 @@ def eval_array_api_strict(self, array_api_strict_obj: Any) -> Any:
             numb_aparam,
             numb_dos,
         ) = self.param
-        return np.asarray(
+        return to_numpy_array(
             array_api_strict_obj(
                 array_api_strict.asarray(self.inputs),
                 array_api_strict.asarray(self.atype.reshape(1, -1)),

source/tests/consistent/fitting/test_ener.py

@@ -6,6 +6,9 @@
 
 import numpy as np
 
+from deepmd.dpmodel.common import (
+    to_numpy_array,
+)
 from deepmd.dpmodel.fitting.ener_fitting import EnergyFittingNet as EnerFittingDP
 from deepmd.env import (
     GLOBAL_NP_FLOAT_PRECISION,
@@ -232,7 +235,6 @@ def eval_jax(self, jax_obj: Any) -> Any:
         )
 
     def eval_array_api_strict(self, array_api_strict_obj: Any) -> Any:
-        array_api_strict.set_array_api_strict_flags(api_version="2023.12")
         (
             resnet_dt,
             precision,
@@ -241,7 +243,7 @@ def eval_array_api_strict(self, array_api_strict_obj: Any) -> Any:
             (numb_aparam, use_aparam_as_mask),
             atom_ener,
         ) = self.param
-        return np.asarray(
+        return to_numpy_array(
             array_api_strict_obj(
                 array_api_strict.asarray(self.inputs),
                 array_api_strict.asarray(self.atype.reshape(1, -1)),

source/tests/consistent/fitting/test_polar.py

@@ -6,6 +6,9 @@
 
 import numpy as np
 
+from deepmd.dpmodel.common import (
+    to_numpy_array,
+)
 from deepmd.dpmodel.fitting.polarizability_fitting import PolarFitting as PolarFittingDP
 from deepmd.env import (
     GLOBAL_NP_FLOAT_PRECISION,
@@ -175,7 +178,7 @@ def eval_jax(self, jax_obj: Any) -> Any:
         )
 
     def eval_array_api_strict(self, array_api_strict_obj: Any) -> Any:
-        return np.asarray(
+        return to_numpy_array(
             array_api_strict_obj(
                 array_api_strict.asarray(self.inputs),
                 array_api_strict.asarray(self.atype.reshape(1, -1)),

source/tests/consistent/fitting/test_property.py

@@ -6,6 +6,9 @@
 
 import numpy as np
 
+from deepmd.dpmodel.common import (
+    to_numpy_array,
+)
 from deepmd.dpmodel.fitting.property_fitting import (
     PropertyFittingNet as PropertyFittingDP,
 )
@@ -226,7 +229,6 @@ def eval_jax(self, jax_obj: Any) -> Any:
         )
 
     def eval_array_api_strict(self, array_api_strict_obj: Any) -> Any:
-        array_api_strict.set_array_api_strict_flags(api_version="2023.12")
         (
             resnet_dt,
             precision,
@@ -236,7 +238,7 @@ def eval_array_api_strict(self, array_api_strict_obj: Any) -> Any:
             task_dim,
             intensive,
         ) = self.param
-        return np.asarray(
+        return to_numpy_array(
             array_api_strict_obj(
                 array_api_strict.asarray(self.inputs),
                 array_api_strict.asarray(self.atype.reshape(1, -1)),

source/tests/consistent/test_activation.py

@@ -7,6 +7,9 @@
 from deepmd.common import (
     VALID_ACTIVATION,
 )
+from deepmd.dpmodel.common import (
+    to_numpy_array,
+)
 from deepmd.dpmodel.utils.network import get_activation_fn as get_activation_fn_dp
 
 from ..seed import (
@@ -21,8 +24,8 @@
 
 if INSTALLED_PT:
     from deepmd.pt.utils.utils import ActivationFn as ActivationFn_pt
+    from deepmd.pt.utils.utils import to_numpy_array as torch_to_numpy
     from deepmd.pt.utils.utils import (
-        to_numpy_array,
         to_torch_tensor,
     )
 if INSTALLED_TF:
@@ -59,7 +62,7 @@ def test_tf_consistent_with_ref(self):
     @unittest.skipUnless(INSTALLED_PT, "PyTorch is not installed")
     def test_pt_consistent_with_ref(self):
         if INSTALLED_PT:
-            test = to_numpy_array(
+            test = torch_to_numpy(
                 ActivationFn_pt(self.activation)(to_torch_tensor(self.random_input))
             )
             np.testing.assert_allclose(self.ref, test, atol=1e-10)
@@ -70,12 +73,9 @@ def test_pt_consistent_with_ref(self):
     def test_arary_api_strict(self):
         import array_api_strict as xp
 
-        xp.set_array_api_strict_flags(
-            api_version=get_activation_fn_dp.array_api_version
-        )
         input = xp.asarray(self.random_input)
         test = get_activation_fn_dp(self.activation)(input)
-        np.testing.assert_allclose(self.ref, np.array(test), atol=1e-10)
+        np.testing.assert_allclose(self.ref, to_numpy_array(test), atol=1e-10)
 
     @unittest.skipUnless(INSTALLED_JAX, "JAX is not installed")
     def test_jax_consistent_with_ref(self):

source/tests/consistent/test_type_embedding.py

@@ -6,6 +6,9 @@
 
 import numpy as np
 
+from deepmd.dpmodel.common import (
+    to_numpy_array,
+)
 from deepmd.dpmodel.utils.type_embed import TypeEmbedNet as TypeEmbedNetDP
 from deepmd.utils.argcheck import (
     type_embedding_args,
@@ -130,7 +133,8 @@ def eval_jax(self, jax_obj: Any) -> Any:
     def eval_array_api_strict(self, array_api_strict_obj: Any) -> Any:
         out = array_api_strict_obj()
         return [
-            np.asarray(x) if hasattr(x, "__array_namespace__") else x for x in (out,)
+            to_numpy_array(x) if hasattr(x, "__array_namespace__") else x
+            for x in (out,)
         ]
 
     def extract_ret(self, ret: Any, backend) -> tuple[np.ndarray, ...]: