Raimondas Galvelis

So, the consensus is to use the API from this PR and the argument part from #134. Could somebody approve and I'll merge?

I'm in favour for the second version. We are already using more structure inputs for the data loaders. Regarding the CLI, we maintain the ability to override simple options like...

@sarthakpati thanks for your effort! Do you have plant to make packages for *PyTorch Geometric* and its other dependencies?

This is obsolted

@anicoli could you share the input file to reproduce the issue?

Just to keep things clean. Garbage tends to accumulate over time.

NumPy has C API. So some header files are part of it.

TorchANI 2.2 or 2.2.2?

The same issue like https://github.com/aiqm/torchani/issues/598. You need TorchANI 2.2.2.

> 2.2 is also the most recent release at https://github.com/aiqm/torchani. They released TorchANI 2.2.2 (https://github.com/conda-forge/torchani-feedstock/pull/3), but not updated https://github.com/aiqm/torchani