Repository Issue Activity (beta)

#685 License clarification for the published RT-DETRv2 pretrained weights

#672 DDP training hangs when `num_denoising > 0` and batch contains zero ground-truth targets

#678 nn.MultiheadAttention is called with the default need_weights=True, which disables the fused attention kernel

#677 pos_embed is kept on CPU and copied to the GPU on every forward

#668 能否提供论文中模型架构图的源visio文件？

#675 Security: request a private channel to report a vulnerability / 安全漏洞:请求私密披露渠道

#673 50系显卡环境

#558 为什么同数据集训练RTDETR（resnet101）模型在ultralytics上训练效果比这RTDETRv1效果差

#670 训练时的指标问题

#667 请问有只在objects365上预训练的权重吗？

#664 Poor Logits when fine tuning RTDETR

#292 一元机场为什么这么便宜?

#651 TensorRT Error When exporting RT-DETR v2 to TensorRT, I encounter the following error:

#157 关于pytorch版本的DT-DETR在评估时的FPS

#299 FPS相关问题

#662 关于模型训练最终都会报一个相同的错误

#660 📋 Documentation Enhancement Suggestion

#311 固定随机种子后每次跑的结果不同

#635 Train images other than 640x640

#277 The validation loss was large and almost unchanged, and the training loss has been decreasing

#71 Small-object detection

#657 转换模型

#484 export 报错，转rknn

#74 Is there a way to export RT-DETR rknn?

#656 from torchvision import datapoints ImportError: cannot import name 'datapoints' from 'torchvision' (C:\Users\13340\.conda\envs\rtdetr_py38\lib\site-packages\torchvision\__init__.py)