Do you have any tutorial or pointers about how to convert matbench samples to torch_geometric Data objects?
For instance, this is what an "input" looks like for matbench_mp_e_form but not sure how to convert that into a pytorch_geometric Data format to train a GNN on it:
Structure Summary
Lattice
abc : 2.81605053 2.81605053 2.81605053
angles : 90.0 90.0 90.0
volume : 22.331676604441288
A : 2.81605053 0.0 -0.0
B : -0.0 2.81605053 -0.0
C : 0.0 0.0 2.81605053
pbc : True True True
PeriodicSite: Pt (0.0000, 0.0000, 0.0000) [0.0000, 0.0000, 0.0000]
PeriodicSite: C (1.4080, 1.4080, 1.4080) [0.5000, 0.5000, 0.5000]
I found this AtomsToGraph class in the OCP project and I want to make sure it does the right thing if I use it like:
from atoms_to_graphs import AtomsToGraph
from pymatgen.io.ase import AseAtomsAdaptor
from matbench.bench import MatbenchBenchmark
mb = MatbenchBenchmark(autoload=False)
task = mb.task_map["matbench_mp_e_form"]
task.load()
train_inputs, train_outputs = task.get_train_and_val_data(0)
struct = train_inputs[0]
atoms = AseAtomsAdaptor.get_atoms(struct)
data = AtomsToGraph(atoms)
Do you have any tutorial or pointers about how to convert
matbenchsamples totorch_geometricDataobjects?For instance, this is what an "input" looks like for
matbench_mp_e_formbut not sure how to convert that into apytorch_geometricDataformat to train a GNN on it:I found this
AtomsToGraphclass in the OCP project and I want to make sure it does the right thing if I use it like: