How can I do the same thing when 'label' is a list? For example: label = [2,4,6,1,7...,9]
label = 3
NumClass = 10
NumRows = 100
mask = torch.zeros(100, 64)
ones = torch.ones(1, 64)
ElementsPerClass = NumRows//NumClass
mask [ ElementsPerClass*label : ElementsPerClass*(label+1) ] = ones
You are looking for scatter
:
NumRows = len(label)
mask = torch.zeros((NumRoes, NumClass)).scatter_(dim=1, index=torch.tensor(label, dtype=torch.long)[:, None], src=torch.ones(NumRows, 1))