RuntimeError: Some elements marked as dirty during the forward method were not returned as output. The inputs that are modified inplace must all be outputs of the Function. when finetune the pretrained model on coco dataset

[huqianqian123]

I want to know how to solve it.
The detailed report is:
Traceback (most recent call last):
File "tools/train_net.py", line 218, in
main()
File "tools/train_net.py", line 211, in main
model = train(cfg, args.local_rank, args.distributed)
File "tools/train_net.py", line 101, in train
do_train(
File "/home/hqq/DetNAS-master/maskrcnn_benchmark/engine/trainer.py", line 116, in do_train
loss_dict = model(images, targets)
File "/home/hqq/miniconda3/envs/py38/lib/python3.8/site-packages/torch/nn/modules/module.py", line 1130, in _call_impl
return forward_call(*input, **kwargs)
File "/home/hqq/miniconda3/envs/py38/lib/python3.8/site-packages/torch/nn/parallel/distributed.py", line 1008, in forward
output = self._run_ddp_forward(*inputs, **kwargs)
File "/home/hqq/miniconda3/envs/py38/lib/python3.8/site-packages/torch/nn/parallel/distributed.py", line 969, in _run_ddp_forward
return module_to_run(*inputs[0], **kwargs[0])
File "/home/hqq/miniconda3/envs/py38/lib/python3.8/site-packages/torch/nn/modules/module.py", line 1130, in _call_impl
return forward_call(*input, **kwargs)
File "/home/hqq/miniconda3/envs/py38/lib/python3.8/site-packages/apex-0.1-py3.8.egg/apex/amp/_initialize.py", line 196, in new_fwd
output = old_fwd(*applier(args, input_caster),
File "/home/hqq/DetNAS-master/maskrcnn_benchmark/modeling/detector/generalized_rcnn.py", line 56, in forward
features = self.backbone(images.tensors)
File "/home/hqq/miniconda3/envs/py38/lib/python3.8/site-packages/torch/nn/modules/module.py", line 1130, in _call_impl
return forward_call(*input, **kwargs)
File "/home/hqq/miniconda3/envs/py38/lib/python3.8/site-packages/torch/nn/modules/container.py", line 139, in forward
input = module(input)
File "/home/hqq/miniconda3/envs/py38/lib/python3.8/site-packages/torch/nn/modules/module.py", line 1130, in _call_impl
return forward_call(*input, **kwargs)
File "/home/hqq/DetNAS-master/maskrcnn_benchmark/modeling/backbone/detnasnet.py", line 68, in forward
x = self.first_conv(x)
File "/home/hqq/miniconda3/envs/py38/lib/python3.8/site-packages/torch/nn/modules/module.py", line 1130, in _call_impl
return forward_call(*input, **kwargs)
File "/home/hqq/DetNAS-master/maskrcnn_benchmark/modeling/backbone/shuffle_blocks.py", line 144, in forward
x = self.bn(x)
File "/home/hqq/miniconda3/envs/py38/lib/python3.8/site-packages/torch/nn/modules/module.py", line 1130, in _call_impl
return forward_call(*input, **kwargs)
File "/home/hqq/DetNAS-master/maskrcnn_benchmark/pytorch_distributed_syncbn/syncbn.py", line 104, in forward
return DistributedSyncBNFucntion.apply(x, self.weight, self.bias, self.running_mean, self.running_var,
RuntimeError: Some elements marked as dirty during the forward method were not returned as output. The inputs that are modified inplace must all be outputs of the Function.