Using tools/train.py file to train the model and test it, you cannot get the result

[Sakura-rain]

I tried to train the model on the HOIW dataset, using Tools /train.py file for training, and got the model, but when using the model to train on tools/ test_hoiW. py, the obtained hoiw_result.json file has no result. As shown in the figure below.

We can get the results by testing with the hoiw_lastest.pth file provided by the author

I tried to continue the training under Hoiw_lastest.pth, but the following problems occurred and the training could not be carried out

---

missing keys in source state_dict: layer2.0.conv2_offset.weight, layer2.0.conv2_offset.bias, layer2.1.conv2_offset.weight, layer2.1.conv2_offset.bias, layer2.2.conv2_offset.weight, layer2.2.conv2_offset.bias, layer2.3.conv2_offset.weight, layer2.3.conv2_offset.bias, layer3.0.conv2_offset.weight, layer3.0.conv2_offset.bias, layer3.1.conv2_offset.weight, layer3.1.conv2_offset.bias, layer3.2.conv2_offset.weight, layer3.2.conv2_offset.bias, layer3.3.conv2_offset.weight, layer3.3.conv2_offset.bias, layer3.4.conv2_offset.weight, layer3.4.conv2_offset.bias, layer3.5.conv2_offset.weight, layer3.5.conv2_offset.bias, layer3.6.conv2_offset.weight, layer3.6.conv2_offset.bias, layer3.7.conv2_offset.weight, layer3.7.conv2_offset.bias, layer3.8.conv2_offset.weight, layer3.8.conv2_offset.bias, layer3.9.conv2_offset.weight, layer3.9.conv2_offset.bias, layer3.10.conv2_offset.weight, layer3.10.conv2_offset.bias, layer3.11.conv2_offset.weight, layer3.11.conv2_offset.bias, layer3.12.conv2_offset.weight, layer3.12.conv2_offset.bias, layer3.13.conv2_offset.weight, layer3.13.conv2_offset.bias, layer3.14.conv2_offset.weight, layer3.14.conv2_offset.bias, layer3.15.conv2_offset.weight, layer3.15.conv2_offset.bias, layer3.16.conv2_offset.weight, layer3.16.conv2_offset.bias, layer3.17.conv2_offset.weight, layer3.17.conv2_offset.bias, layer3.18.conv2_offset.weight, layer3.18.conv2_offset.bias, layer3.19.conv2_offset.weight, layer3.19.conv2_offset.bias, layer3.20.conv2_offset.weight, layer3.20.conv2_offset.bias, layer3.21.conv2_offset.weight, layer3.21.conv2_offset.bias, layer3.22.conv2_offset.weight, layer3.22.conv2_offset.bias, layer4.0.conv2_offset.weight, layer4.0.conv2_offset.bias, layer4.1.conv2_offset.weight, layer4.1.conv2_offset.bias, layer4.2.conv2_offset.weight, layer4.2.conv2_offset.bias

None
loading annotations into memory...
Done (t=0.41s)
creating index...
index created!
2022-09-09 01:38:15,562 - INFO - load checkpoint from hoiw_latest.pth
2022-09-09 01:38:17,626 - WARNING - The model and loaded state dict do not match exactly

missing keys in source state_dict: sa.g.conv.weight, sa.g.conv.bias, sa.theta.conv.weight, sa.theta.conv.bias, sa.phi.conv.weight, sa.phi.conv.bias, sa.conv_out.conv.weight, sa.conv_out.conv.bias

Traceback (most recent call last):
File "tools/train.py", line 108, in
main()
File "tools/train.py", line 104, in main
logger=logger)
File "/root/anaconda3/envs/py37_torch/lib/python3.7/site-packages/mmdet-1.0rc0+65c1842-py3.7-linux-x86_64.egg/mmdet/apis/train.py", line 62, in train_detector
_non_dist_train(model, dataset, cfg, validate=validate)
File "/root/anaconda3/envs/py37_torch/lib/python3.7/site-packages/mmdet-1.0rc0+65c1842-py3.7-linux-x86_64.egg/mmdet/apis/train.py", line 287, in _non_dist_train
runner.resume(cfg.resume_from)
File "/root/anaconda3/envs/py37_torch/lib/python3.7/site-packages/mmcv/runner/runner.py", line 314, in resume
self.optimizer.load_state_dict(checkpoint['optimizer'])
File "/root/anaconda3/envs/py37_torch/lib/python3.7/site-packages/torch/optim/optimizer.py", line 141, in load_state_dict
raise ValueError("loaded state dict has a different number of "
ValueError: loaded state dict has a different number of parameter groups

---

In addition, missing Keys will also appear when the model provided by the author is used for testing, while the model trained by the train.py file will not appear:

---

The model and loaded state dict do not match exactly:

missing keys in source state_dict: sa.g.conv.weight, sa.g.conv.bias, sa.theta.conv.weight, sa.theta.conv.bias, sa.phi.conv.weight, sa.phi.conv.bias, sa.conv_out.conv.weight, sa.conv_out.conv.bias

---

The GPU I use is V100
python=3.7
mmcv=0.4.3
mmdet =1.0rc0+65c1842
pytorch= 1.10

I hope the author can answer.Thank you.