Pretrained weights such as [Faster RCNN + GIoU] you supported can't use

[yunxi1]

Hi,I download your trained model faster_giou.pth(333mb)，and run
python tools/test_net.py --dataset coco2017 --cfg configs/baselines/e2e_faster_rcnn_R-50-FPN_giou_1x.yaml --load_ckpt
faster_giou.pth
ERROR:
RuntimeError: Error(s) in loading state_dict for Generalized_RCNN:
While copying the parameter named "Box_Head.fc1.weight", whose dimensions in the model are torch.Size([1024, 50176]) and whose dimensions in the checkpoint are torch.Size([1024, 12544]).

Then I run
python tools/train_net_step.py --dataset coco2017 --cfg configs/baselines/e2e_faster_rcnn_R-50-FPN_giou_1x.yaml --use_tfboard
and get my model(641mb),
so I confused which model is right ? 641MB vs 333MB?
I need your help,thanks