search

kmzzhang / nbi

Package for Neural Posterior Estimation and Importance Sampling focused on Astronomical Applications
Other
35 stars 4 forks source link

Support multi channel input #14

Open kmzzhang opened 1 year ago

kmzzhang commented 1 year ago

The featurizer input needs to support a list of featurizer configs NBI also need to support simulator outputs of type list where each element is fed into the corresponding featurizer

jackieblaum commented 1 year ago

class multi(nn.Module): def init(self.list_of_nets)" self.nets = list_of_nets self.n_net = len(list_of_nets) def forward(self, list_of_x): out = list() for i in range(self.n_net): out.append(self.neti) return torch.cat(out)

Also in nbi.engine the functions scale_x and should also allow for multi channel input

kz