AttributeError: 'RPN' object has no attribute 'edge_conv2d' when training with rpn_r50_fpn_1x_coco.py on COCO dataset

[chan1031]

Hi,

I'm trying to train a model on the COCO dataset using the rpn_r50_fpn_1x_coco.py configuration with MMDetection. However, I'm encountering an AttributeError related to the edge_conv2d attribute in the RPN class. Below is the error traceback:

creating index...
index created!
loading annotations into memory...
Done (t=0.61s)
creating index...
index created!
09/24 00:26:27 - mmengine - INFO - load model from: torchvision://resnet50
09/24 00:26:27 - mmengine - INFO - Loads checkpoint by torchvision backend from path: torchvision://resnet50
09/24 00:26:27 - mmengine - WARNING - The model and loaded state dict do not match exactly

unexpected key in source state_dict: fc.weight, fc.bias

09/24 00:26:27 - mmengine - INFO - Checkpoints will be saved to /home/skku/mmdetection/work_dirs/rpn_r50_fpn_1x_coco.
Traceback (most recent call last):
  ...
  File "/home/skku/mmdetection/mmdet/models/detectors/single_stage.py", line 154, in extract_feat
    x = self.edge_conv2d(batch_inputs)
AttributeError: 'RPN' object has no attribute 'edge_conv2d'

Steps to Reproduce:
Use the rpn_r50_fpn_1x_coco.py configuration file.
Start training the model on the COCO dataset.

Environment:
MMDetection version: 3.30

Additional Information:
I have not modified the RPN class directly, and I'm unsure why edge_conv2d is being referenced. Could this be related to an incorrect or outdated configuration, or is it a potential bug?

Thanks!