Bugfix: Interactive calculation array comparison #1542
Conversation
This seems to related to the recent numpy updates
for more information, see https://pre-commit.ci
| elif np.array(data).dtype == np.dtype("O"): | ||
| self._extend_hdf(h5=h5, path=path, key=key, data=data) | ||
| else: | ||
| self._extend_hdf(h5=h5, path=path, key=key, data=np.array(data)) |
There was a problem hiding this comment.
All these calls to the same function are very similar. Actually, why can the stuff from the except clause not be done here? If not a dtype=O, then construct such an array instead of a regular one?
There was a problem hiding this comment.
With the recent numpy version np.array(data) fails, unless np.array(data, dtype="object") is set.
There was a problem hiding this comment.
This probably means that this clause elif np.array(data).dtype == np.dtype("O") is never trigger, because it raises an exception.
There was a problem hiding this comment.
I agree, but I am not sure. The clean way would require the user to specify if the number of atoms is going to change in their simulation or not and then choose the corresponding approach for storing the structures. So this is really just a quick fix to use interactive jobs to evaluate machine learned interatomic potentials to calculate the energies and forces for the test and training set.
This seems to related to the recent numpy updates