BigDong

seems the ckpt have some problem, we will fix it ASAP

| Model | backbone | AR@1000 | |:------:|:------:|:------:| | RPN-FPN | resnet50 | 58.30 | | RPN-C4 | resnet50-caffe | 51.28 | The proposals are converted from detectron2 ([train](https://dl.fbaipublicfiles.com/detectron2/COCO-Detection/rpn_R_50_FPN_1x/137258492/coco_2017_train_box_proposals_21bc3a.pkl) an...

@YvesXu Thanks for your contribution, please fix the comments and train the CondInst model to make sure the results is same as the original repo

Haven't met this error before, maybe you can refer to https://github.com/microsoft/vscode-jupyter/wiki/Kernel-crashes-when-using-numpy

Seems your config file is normal. Did your runner have a for loop? did not see it in your issue

For WIDER FACE dataset, you can refer to https://github.com/open-mmlab/mmdetection/pull/8508

So, this seems that your training or testing loop have some problem

maybe you can refer to https://github.com/open-mmlab/mmdetection/blob/3b72b12fe9b14de906d1363982b9fba05e7d47c1/mmdet/apis/test.py#L104-L106

`r('"file" must be a filename str or a file-object')` please check your dumped file name

This PR is to fix errors in lower version PyTorch, which will not affect high version PyTorch