Document that Molecule.from_dict accepts a list of lists but not a numpy array #1998
Assignees
Labels
Comments
|
You probably already figured it out, but the fulcrum of the issue is what types are handled by the little helper >>> from openff.toolkit.utils.utils import deserialize_numpy
/Users/mattthompson/micromamba/envs/openff-interchange-env/lib/python3.11/site-packages/openff/utilities/provenance.py:23: CondaExecutableNotFoundWarning: No conda/mamba/micromamba executable found. Unable to determine package versions.
warnings.warn(
>>> import numpy
>>> deserialize_numpy(numpy.eye(3).tolist(), shape=(3, 3))
array([[1., 0., 0.],
[0., 1., 0.],
[0., 0., 1.]])
>>> deserialize_numpy(numpy.eye(3), shape=(3, 3))
Traceback (most recent call last):
File "<stdin>", line 1, in <module>
File "/Users/mattthompson/software/openff-toolkit/openff/toolkit/utils/utils.py", line 433, in deserialize_numpy
np_array = np_array.reshape(shape)
^^^^^^^^
UnboundLocalError: cannot access local variable 'np_array' where it is not associated with a value
>>>Something like this should smooth that over:
diff --git a/openff/toolkit/utils/utils.py b/openff/toolkit/utils/utils.py
index db599d0a..1f64e80a 100644
--- a/openff/toolkit/utils/utils.py
+++ b/openff/toolkit/utils/utils.py
@@ -406,7 +406,7 @@ def serialize_numpy(np_array) -> tuple[bytes, tuple[int]]:
def deserialize_numpy(
- serialized_np: Union[bytes, list],
+ serialized_np: Union[NDArray, bytes, list],
shape: tuple[int, ...],
) -> NDArray:
"""
@@ -425,11 +425,19 @@ def deserialize_numpy(
np_array
The deserialized numpy array
"""
- if isinstance(serialized_np, list):
+ if isinstance(serialized_np, np.ndarray):
+ np_array = serialized_np
+ elif isinstance(serialized_np, list):
np_array = np.array(serialized_np)
- if isinstance(serialized_np, bytes):
+ elif isinstance(serialized_np, bytes):
dt = np.dtype("float").newbyteorder(">")
np_array = np.frombuffer(serialized_np, dtype=dt)
+ else:
+ raise ValueError(
+ f"Unexpected type ({type(serialized_np)}) found when attempting to deserialize from the "
+ "serialized form of a numpy array."
+ )
+
np_array = np_array.reshape(shape)
return np_array
>>> deserialize_numpy(numpy.eye(3).tolist(), shape=(3, 3))
array([[1., 0., 0.],
[0., 1., 0.],
[0., 0., 1.]])
>>> deserialize_numpy(numpy.eye(3), shape=(3, 3))
array([[1., 0., 0.],
[0., 1., 0.],
[0., 0., 1.]])
>>>Arguably this is unsupported since the dict model of a |
|
Yeah, Matt's right here - This isn't intended to be a supported use case. This might be resolved by better docs (like, "the exact format of the input dict for this method is determined by the output of Molecule.to_dict") |
j-wags
changed the title
5 tasks
Sign up for free
to join this conversation on GitHub.
Already have an account?
Sign in to comment
Describe the bug
When using the
Molecule.from_dictmethod,molecule_dict["conformers"][0]cannot be a numpy array.To Reproduce
Output
Computing environment (please complete the following information):
conda list:Additional context
removing np.array() from the conformers key (so a list of lists) will allow the example to run as expected.
The text was updated successfully, but these errors were encountered: