input data length about boda2 HOT 1 CLOSED

@wawpaopao, we used weights from Basset, trained on DHS data from >100 cell lines, to initialize layers in our model that matched. We found this was very helpful for training models on our MPRA data set. The downside was that Basset was a fixed input length model (600 nt) so our model maintained that requirement, despite the training inputs being shorter (<=200 nt). We decided to pad the elements with the flanking sequences from the MPRA vector that reflects the local sequence context of the MPRA oligos.

Apologies for the delay in my response. HTH.

from boda2.