Preparedata.py dividing w_forces_column with float error and contradiction with custom weights notebook

[Nick19-bot]

Hi!
I've been trying to incorporate weights to the training of my Cu dataset to try and replicate the potential of a paper, during this i've been doing it all without the weights that are provided in the dataset and recently i've started to incorporate but have had some problems in the process. I've extracted the w_energy and w_forces from my 50k structures dataset and i'm trying to debug it with the first 25 structures for easy management.

The first error i encountered is one that says the indexes of the weights dataset are not on the original dataset, but as far as i checked the indices match up with the raw dataset that i was provided with and i don't know if there's some sort of distinction pandas does that raises this error.
I've checked the structures that pair up nicely with their weights counterparts and i can't discern a difference between the problematic ones.
Removing the structures and rerunning the program seems to fix the issue but i'd like to preserve as many structures as possible in order to get an optimally trained potential.

This brings me to the second error i encounter once i remove the problematic structures, apparently preparedata.py is dividing the weight's data-frame forces column with the w_forces norm but the column is a collection of lists and it raises an TypeError, i looked this up in the weights usage notebook and the structure that is used is the same as my data-frames at first glance, the w_forces_column is a collection of lists with entries for each atom of each structure, so i don't see any clear differences from my dataset.

If anyone has any feedback on what may be happening or how can i check for differences in the working and problematic structures, and the true format of the w_forces_column i would very much appreciate it.

Kind regards,
Nick