model feature extraction not possible on cuda
florianmahner opened this issue · 1 comments
florianmahner commented
if torch.device is cuda, then thingsvision.model_class.Model.extract_features()
throws a type error, since features are torch.Tensor()
and features = np.vstack(features)
#L223 is not possible. suggested fix on #L221 features.append(act.cpu().numpy())
LukasMut commented
Feel free to open a PR. :)