Is your feature request related to a problem? Please describe.
TimeSeriesCondition only supports torch.Tensor/LabelTensor input. There is no equivalent for graph-structured data (Data/Graph), forcing users to manually pre-process temporal windows when working with graph neural operators.
Describe the solution you'd like
A GraphTimeSeriesCondition extending TimeSeriesCondition for torch_geometric.data.Data and pina.Graph inputs, with the same windowing and autoregressive rollout logic, configurable via a node-feature key (e.g. "x").
Describe alternatives you've considered
Using DataCondition with pre-windowed snapshots — works but moves windowing logic outside the library.
Additional context
/
Is your feature request related to a problem? Please describe.
TimeSeriesCondition only supports torch.Tensor/LabelTensor input. There is no equivalent for graph-structured data (Data/Graph), forcing users to manually pre-process temporal windows when working with graph neural operators.
Describe the solution you'd like
A GraphTimeSeriesCondition extending TimeSeriesCondition for torch_geometric.data.Data and pina.Graph inputs, with the same windowing and autoregressive rollout logic, configurable via a node-feature key (e.g. "x").
Describe alternatives you've considered
Using DataCondition with pre-windowed snapshots — works but moves windowing logic outside the library.
Additional context
/