Error importing Open Catalyst 2022 LMDB files

[allaffa]

Python version

Python 3.11.5

fairchem-core version

1.2.1

pytorch version

2.6.0

cuda version

12.1

Operating system version

Linux

Minimal example

def traj_to_torch_geom(self, traj_file):
        # Open LMDB
        env = lmdb.open(traj_file, subdir=False, readonly=True, lock=False, readahead=False, meminit=False)

        with env.begin() as txn:
            cursor = txn.cursor()

            for key, value in iterate_tqdm(cursor, verbosity_level=2, desc="Processing OC22 LMDB"):
                old_data = pickle.loads(value)  # Load trajectory data
                print(old_data)

Current behavior

When I import the LMDB files of the Open Catalyst 2022 dataset and try to load the PyG data objects, I obtain the following error

RuntimeError: The 'data' object was created by an older version of PyG. If this error occurred while loading an already existing dataset, remove the 'processed/' directory in the dataset's root folder and try again.

To my understanding, Open Catalyst 2022 was previously released in .traj format, and this enabled a more flexible import that was not strongly dependent off the version of the packages.
Is there any way that:

1. the incompatibility between PyG versions can be solved - No, I do not want to downgrade my version of PyG
2. make available the raw data of Open Catalyst 2022, which allows for more flexibility on the user?

Expected Behavior

I would like the code not to complain about versions of PyG when importing Data objects

Releasing the dataset in a format the enforces using a specific version of PyG severely affects the usability of this dataset. Providing the raw output in XYZ formats would enable a much wider usage of the dataset.

Relevant files to reproduce this bug

No response