Closed bvhari closed 3 months ago
Without the change, drop_rate=0 will cause the creation of an IntTensor from torch.full, which causes torch.bernoulli to throw an error as it expects a FloatTensor
Without the change, drop_rate=0 will cause the creation of an IntTensor from torch.full, which causes torch.bernoulli to throw an error as it expects a FloatTensor