npz training dataset has different cells in PBC

[afour9961207]

Hello,

I'm considering training with the PBC dataset, but my cells may vary.
According to previous discussion, if my training data doesn't have the same cell, I should place the PBC and cell tags in the key_mapping (not in npz_fixed_field_keys).
They should have pbc = (True, True, True) in the total shape of (nframe, 3) and cell = [[19.57, 0, 0], [0, 19.57, 0], [0, 0, 19.57]] in the total shape of (nframe, 3, 3).

Am I misunderstanding anything, or is this a correct setup?

Thanks for your clarification!

[Linux-cpp-lisp]

If I understand right this is fine, but you have to map to the actual key name (pbc or cell), not this plain english description of the field.