Yajie Wang

RuntimeError: All input tensors must be on the same device. Received cuda:0 and cuda:3

3

File "./train.py", line 189, soft_preds = torch.cat(soft_parsing, dim = 0). Does anyone meet this error?