We want to provide a dataset structure for the most common datasets that are provided through the QCArchive infrastructure. These datasets should play nicely with the pytorchdataloader objects and inherit from torch.utils.data.Dataset.
Todos
[x] Outline general structure
[x] Implement QM9 toy dataset
[ ] Tests are passing
Questions
[ ] How to best query different labels?
[ ] Is there any use case for different datapoints beside coordinates?
[ ] Is a view the best structure to perform queries on?
Description
We want to provide a dataset structure for the most common datasets that are provided through the QCArchive infrastructure. These datasets should play nicely with the
pytorch
dataloader
objects and inherit fromtorch.utils.data.Dataset
.Todos
Questions
view
the best structure to perform queries on?Status