The H deformable detr code may have some problems !

[Cao-PeiPei]

I was able to train with other code, but training with H deformable detr code with the same configuration would cause errors. Can you update it on github? Thankyou！！！！！！！The error is as follows:
Traceback (most recent call last):
File "detrex/tools/train_net.py", line 307, in
launch(
File "/root/detrex/detectron2/detectron2/engine/launch.py", line 82, in launch
main_func(*args)
File "detrex/tools/train_net.py", line 302, in main
do_train(args, cfg)
File "detrex/tools/train_net.py", line 275, in do_train
trainer.train(start_iter, cfg.train.max_iter)
File "/root/detrex/detectron2/detectron2/engine/train_loop.py", line 149, in train
self.run_step()
File "detrex/tools/train_net.py", line 101, in run_step
loss_dict = self.model(data)
File "/root/miniconda3/lib/python3.8/site-packages/torch/nn/modules/module.py", line 1102, in _call_impl
return forward_call(*input, **kwargs)
File "/root/detrex/projects/h_deformable_detr/modeling/h_deformable_detr.py", line 231, in forward
) = self.transformer(
File "/root/miniconda3/lib/python3.8/site-packages/torch/nn/modules/module.py", line 1102, in _call_impl
return forward_call(*input, **kwargs)
File "/root/detrex/projects/h_deformable_detr/modeling/h_deformable_transformer.py", line 462, in forward
inter_states, inter_references = self.decoder(
File "/root/miniconda3/lib/python3.8/site-packages/torch/nn/modules/module.py", line 1102, in _call_impl
return forward_call(*input, **kwargs)
File "/root/detrex/projects/h_deformable_detr/modeling/h_deformable_transformer.py", line 211, in forward
new_reference_points if self.look_forward_twice else reference_points
UnboundLocalError: local variable 'new_reference_points' referenced before assignment

[rentainhe]

Hello, we will check the problem and fix the bug as soon as possible. @Cao-PeiPei

[Cao-PeiPei]

Thankyou！！！！！！！！

[Cao-PeiPei]

@rentainhe Excuse me, I tried it and the above problem still appears, did you change the code? If the code is modified, I'll try again myself and look forward to your reply.