DK-DETR training problem

[Lukas-Ma1]

Thank you for your dkdetr , I use the command: CUDA_VISIBLE_DEVICES=0 bash tools/dist_train.sh configs/dk-detr/dkd_r50_70e_lvis.py 1 --work-dir ./train_dir to train dk-detr，and there occurs a mistake:
Traceback (most recent call last):
File "tools/train.py", line 242, in
main()
File "tools/train.py", line 238, in main
meta=meta)
File "/data/msy/opera/opera/apis/train.py", line 245, in train_model
runner.run(data_loaders, cfg.workflow)
File "/data/msy/opera/third_party/mmcv/mmcv/runner/epoch_based_runner.py", line 130, in run
epoch_runner(data_loaders[i], **kwargs)
File "/data/msy/opera/third_party/mmcv/mmcv/runner/epoch_based_runner.py", line 51, in train
self.run_iter(data_batch, train_mode=True, **kwargs)
File "/data/msy/opera/third_party/mmcv/mmcv/runner/epoch_based_runner.py", line 30, in run_iter
**kwargs)
File "/data/msy/opera/third_party/mmcv/mmcv/parallel/distributed.py", line 59, in train_step
output = self.module.train_step(*inputs[0], **kwargs[0])
File "/data/msy/opera/third_party/mmdetection/mmdet/models/detectors/base.py", line 248, in train_step
losses = self(**data)
File "/home/xd/anaconda3/envs/dkdetr/lib/python3.7/site-packages/torch/nn/modules/module.py", line 889, in _call_impl
result = self.forward(*input, **kwargs)
File "/data/msy/opera/third_party/mmcv/mmcv/runner/fp16_utils.py", line 116, in new_func
return old_func(*args, **kwargs)
File "/data/msy/opera/third_party/mmdetection/mmdet/models/detectors/base.py", line 172, in forward
return self.forward_train(img, img_metas, **kwargs)
File "/data/msy/opera/third_party/mmdetection/mmdet/models/detectors/single_stage.py", line 84, in forward_train
gt_labels, gt_bboxes_ignore)
File "/data/msy/opera/opera/models/dense_heads/soit_head.py", line 159, in forward_train
outs = self(x, img_metas)
File "/home/xd/anaconda3/envs/dkdetr/lib/python3.7/site-packages/torch/nn/modules/module.py", line 889, in _call_impl
result = self.forward(*input, **kwargs)
TypeError: forward() missing 1 required positional argument: 'text_feats'

Why it notes me 'text_feats' is missed, I include text feat path in config file.

[Lukas-Ma1]

After this, can you give me an example of training command? I have some idea and I would like to modify your code. Thank you so much!

[Letitialky]

@Lukas-Ma1 I run into the same problem, did you solve it ? If so, can you share the solution? Thank you!