sequence numerical stats computation
weihua916 opened this issue · comments
Weihua Hu commented
Currently, sequence numerical stats will be NaN as long as we have a single sequence with NaN element. This is because dropna()
in https://github.com/pyg-team/pytorch-frame/blob/master/torch_frame/data/stats.py#L101 cannot drop NaN elements inside the sequence.
We want to handle NaN in sequence gracefully and augment the test case here.