The IndexError interrupts training

[Ternura111]

Traceback (most recent call last):
File "/VisCom-SSD-2/hj/RVOS/DsHmp/dshmp/engine/train_loop.py", line 149, in train
self.run_step()
File "/VisCom-SSD-2/hj/RVOS/DsHmp/dshmp/engine/defaults.py", line 494, in run_step
self._trainer.run_step()
File "/VisCom-SSD-2/hj/RVOS/DsHmp/dshmp/engine/train_loop.py", line 395, in run_step
loss_dict = self.model(data, self.iter)
File "/home/ta/anaconda3/envs/ylf_rvos/lib/python3.9/site-packages/torch/nn/modules/module.py", line 1110, in _call_impl
return forward_call(*input, **kwargs)
File "/home/ta/anaconda3/envs/ylf_rvos/lib/python3.9/site-packages/torch/nn/parallel/distributed.py", line 963, in forward
output = self.module(*inputs[0], **kwargs[0])
File "/home/ta/anaconda3/envs/ylf_rvos/lib/python3.9/site-packages/torch/nn/modules/module.py", line 1110, in _call_impl
return forward_call(*input, **kwargs)
File "/VisCom-SSD-2/hj/RVOS/DsHmp/dshmp/dshmp_model.py", line 288, in forward
return self.train_model(batched_inputs, iterations)
File "/VisCom-SSD-2/hj/RVOS/DsHmp/dshmp/dshmp_model.py", line 321, in train_model
motion_feat = torch.cat([lang_feat_fusion[motion_map.bool()], lang_feat], dim=0)
IndexError: The shape of the mask [1, 40] at index 0 does not match the shape of the indexed tensor [2, 40, 256] at index 0
[09/09 15:46:29 d2.utils.events]: iter: 0 lr: N/A max_mem: 1355M

[heshuting555]

Thank you for your interest in our work!

Current code base only support one GPU one sample setting! Thanks!

[Sjunshu]

how can i use more than one card，i also meet this problem

[Ternura111]

我怎么能用多张卡，我也遇到这个问题

When training , make sure to use 8 gpu . This works .

[Sjunshu]

Is there a way to train with fewer than 8 GPUs but more than 1 GPU?

[Sjunshu]

I just have 6 GPUs

[Ternura111]

I just have 6 GPUs

sorry , I have no ideas . Maybe the author has a way

[Sjunshu]

OK,thank you