RuntimeError: grad can be implicitly created only for scalar outputs

[whuhxb]

Hi @haotian-liu

Have you ever met this bug before? Thanks.

2023-03-16 18:37:25,684 - pretrain_shapenet - INFO - Using Data parallel ...
Traceback (most recent call last):
File "main.py", line 116, in
main()
File "main.py", line 112, in main
pretrain(args, config, train_writer, val_writer)
File "/export/home/hanxiaobing/Documents/PlaneNet_PlaneRCNN/DGCNN_PointNet2/SensatUrban/MAE/MaskPoint/tools/runner_pretrain.py", line 144, in run_net
loss.backward()
File "/export/home/hanxiaobing/anaconda3/envs/maskpoint/lib/python3.7/site-packages/torch/tensor.py", line 245, in backward
torch.autograd.backward(self, gradient, retain_graph, create_graph, inputs=inputs)
File "/export/home/hanxiaobing/anaconda3/envs/maskpoint/lib/python3.7/site-packages/torch/autograd/init.py", line 141, in backward
grad_tensors = make_grads(tensors, grad_tensors)
File "/export/home/hanxiaobing/anaconda3/envs/maskpoint/lib/python3.7/site-packages/torch/autograd/init.py", line 50, in _make_grads
raise RuntimeError("grad can be implicitly created only for scalar outputs")
RuntimeError: grad can be implicitly created only for scalar outputs

If I revise the _loss.backward() with _loss.backward(torch.ones_like(_loss)), the bugs show like this.

Traceback (most recent call last):
File "main.py", line 116, in
main()
File "main.py", line 112, in main
pretrain(args, config, train_writer, val_writer)
File "/export/home/hanxiaobing/Documents/PlaneNet_PlaneRCNN/DGCNN_PointNet2/SensatUrban/MAE/MaskPoint/tools/runner_pretrain.py", line 168, in run_net
losses.update([loss_1.item(), loss_2.item()])
ValueError: only one element tensors can be converted to Python scalars

Do you have any suggestions to solve this bug？ Thanks a lot.