Zhuofan Zong
Zhuofan Zong
Have you modified the training script? `Bad substitution` means there are some errors in `dist_train.sh` and you can check it.
> Which file to use load_from loading the weight file The image you post shows you are loading the weights from `weights/co_deformable_detr-r50_1.pth`. Please use `load_from` to load it.
Here is an example of MMDet inference: Test Mask R-CNN on COCO test-dev with 8 GPUs, and generate JSON files for submitting to the official evaluation server. Config and checkpoint...
Please first uninstall your mmdetection and reinstall it in the co-detr repo: `pip install -e .`
Sorry for the late reply. The version of the torch we use is `1.8.1+cuda112_cudnn8.1.0`.
应该是单卡训练的config里指定了ddp的port，可以看下是不是这个问题
@onnx20 现在的code可以吗，我更推荐使用COCO的config训练
Co-DETR with LSJ augmentation adopts copy-paste technique, which requires mask segmentation annotation. Please refer to this link for more details: https://github.com/Sense-X/Co-DETR/blob/bf3d49d7c02929788dfe2f251b6b01cbe196b736/projects/configs/co_dino/co_dino_5scale_lsj_r50_1x_coco.py#L68
You can try to use another port instead of 29500. On Thu, Jun 27, 2024 at 4:51 PM Zhuofan Zong ***@***.***> wrote: > > > On Thu, Jun 27, 2024...
Hi, @zhangchbin, we have no plan to do this now. But we have updated the arxiv paper to release more details about this large model.