Mayuchen

We don't support these works in detection area now. And I suggest you to modify our code directly, and I believe it is not difficult to implement these works.

> https://xiulian.blog.csdn.net/article/details/111035882 It works well on 2080Ti and V100. Maybe you should follow the methods as follows: https://blog.csdn.net/m0_38007695/article/details/107065617

> error： when train in single GPU, Default process group is not initialized > solution：https://blog.csdn.net/m0_37568067/article/details/109785209 yeah, it is a well known bug when training on single GPU. We will fix...

Thank you for your attention. https://github.com/Megvii-BaseDetection/cvpods/blob/master/cvpods/layers/border_align.py https://github.com/Megvii-BaseDetection/cvpods/tree/master/cvpods/layers/csrc/border_align

oh, I have found the reason. The loss is nan, owing to the initialization. So I use your pre-trained model. But in your MaskRCNN-benchmark, the pre-trained model for Faster-RCNN and...