Closed josuelandinez closed 1 year ago
Hi @josuelandinez,
Thanks!
I see that CI is failing with TypeError
, so I think that your changes break the compatiibillity. We may need to treat 2 cases separately. It would be good to now when and how it happens that cell matrix is string and not a list/ndarray.
Perhaps @kousuke-nakano would have a better idea.
Hi @josuelandinez,
Thanks!! Would you give me a PySCF checkpoint file to reproduce the bug you reported?
Hi @q-posev and @kousuke-nakano ,
@q-posev you were right, it is my bad habit to parse the cell with strings as Pyscf allows. Then the changes were breaking compatibility. I just change it to treat the different cases as you suggested and seems to pass all the tests now. I also test my self by passing the cell as a list, np.array and a string as I was doing and now the conversion is done in any of these cases, checkpoint files attached.
I think, it can be merged now, I will be attentive to your answer.
best,
Edgar checkpoints.zip
Thanks @josuelandinez ! Good that you catched this corner case. I'll wait for @kousuke-nakano feedback but the PR looks good to me.
Dear @josuelandinez and @q-posev Thanks!! I totally agree to this change :-)
Hi,
I found a problem long time ago when using the converter from pyscf to trexio for periodic systems. I just forget to do the request always. It happens when reading the cell matrix sometimes seems to be just a string and it is not converted properly to a numpy array. The error message is like this:
error:" TypeError: ufunc 'divide' not supported for the input types, and the inputs could not be safely coerced to any supported types according to the casting rule ''safe'' "
I solved in a simple way with the modifications on this fork. I don't think it generates any conflict or other problem. I would like to ask to @kousuke-nakano or @q-posev to review it before merged if you regard it appropriate.
best,
Edgar