I train detectron2 retinanet, get head.cls_score.weight shapes error

[zbsean]

'head.cls_score.weight' to the model due to incompatible shapes: (720, 256, 3, 3) in the checkpoint but (342, 256, 3, 3) in the model! You might want to double check if this is expected.

'head.cls_score.bias' to the model due to incompatible shapes: (720,) in the checkpoint but (342,) in the model! You might want to double check if this is expected.