total energy not found error?

[Mofahdi]

I used the same settings as "config_finetuning.yaml" with one minor change shown below:

data:
target: nequip.data.datamodule.ASEDataModule
seed: 456 # dataset seed for reproducibility
#ase_args: {'total_energy': REF_energy, 'forces': REF_forces}
key_mapping:
total_energy: REF_energy
forces: REF_forces

in my "xyz" file, energy and forces use the keys: REF_energy and REF_forces, respectively

I got the following error:
normed = torch.einsum("n..., n -> n...", data[self.field], num_atoms)
~~~~^^^^^^^^^^^^
KeyError: 'total_energy'

can you please suggest how to solve this error?

[Mofahdi]

one more thing I tried:
I used the same data and script config_finetuning.yaml as used in the tutorial and installed same versions of the packages torch, cuda, and nequip except python was 3.12.13 in my case instead of 3.12.12 shown in the tutorial which i dont think is significant. that was done to reproduce the results. however, i still got an error.
Error executing job with overrides: []
Traceback (most recent call last):

.
.
.
.
.
.
assert isinstance(datamodule, NequIPDataModule)
^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^
AssertionError

[cw-tan]

See docs: https://nequip.readthedocs.io/en/latest/api/datamodule.html#nequip.data.datamodule.ASEDataModule

key_mapping (Dict[str, str]) – mapping of ase keys to AtomicDataDict keys

I think you'd want something like

key_mapping:
  REF_energy: total_energy
  REF_forces: forces

Can you send a more complete stacktrace for

assert isinstance(datamodule, NequIPDataModule)
^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^

I think it's unlikely to be due to python versions. I'm guessing a typo or improper indentation/formatting etc of the config file.