Bug in feature vector access?

[rviramontesTT]

For the below line, is it trying to access the input_cap feature? Since this is the combined features, this would be index 7 (instead of 2 if for just the electrical features)?

MLBuf_MLCAD/models/losses.py

Line 186 in 22971f6

child_cap_k = (cluster_probs * features[:, 2].unsqueeze(-1)).sum(dim=0) # [K]

[rviramontesTT]

Actually, on further review, I'm thinking it should be index 8.

I think there is a similar issue with accessing the locations in the feature vector for looking up locations,

MLBuf_MLCAD/models/losses.py

Line 178 in 22971f6

dist_mat = manhattan_dist_matrix(features[:, :2], cluster_buf_loc) # [N,K]

MLBuf_MLCAD/models/losses.py

Line 200 in 22971f6

sink_driver_dist = torch.abs(features[:, :2] - driver_xy).sum(dim=-1) # [N]

should be 3:5 instead of :2?