set randomness= dict(seed=138011772,deterministic=True) in config file, RuntimeError in anchor_head.py

[ttt0666]

When I train Faster-rcnn with the following setting in config file:

randomness= dict(seed=138011772,deterministic=True)

I get the following error:

  File "/mmdetection-main/mmdet/models/dense_heads/anchor_head.py", line 505, in loss_by_feat
    cls_reg_targets = self.get_targets(
  File "/mmdetection-main/mmdet/models/dense_heads/anchor_head.py", line 381, in get_targets
    results = multi_apply(
  File "/mmdetection-main/mmdet/models/utils/misc.py", line 219, in multi_apply
    return tuple(map(list, zip(*map_results)))
  File "/mmdetection-main/mmdet/models/dense_heads/anchor_head.py", line 286, in _get_targets_single
    bbox_weights[pos_inds, :] = 1.0
RuntimeError: linearIndex.numel()*sliceSize*nElemBefore == expandedValue.numel() INTERNAL ASSERT FAILED at "../aten/src/ATen/native/cuda/Indexing.cu":389, please report a bug to PyTorch. number of flattened indices did not match number of elements in the value tensor: 12 vs 3

and I try to change

bbox_weights[pos_inds, :] = 1.0

in anchor_head.py to

bbox_weights[pos_inds, :] = 1 * torch.ones((len(pos_inds), bbox_weights.size(1)), device=bbox_weights.device)

the error disappears. Is this related to the randomness?

Also, I remove the deterministic=True as:

randomness= dict(seed=138011772)

the error also disappears.