DaBaiTuu

I saw someone uploaded svd-f16.safetentors version,though it is not offical,maybe you can try

And To set lowvram = True will really help,but just for scripts which import the streamlit_helpers.py, unfortunately SVD-series are not included

感谢 pin 出来

> 会尽快更新分割的model和config 好的感谢

> 使用自己的训练数据集，转换成coco2017格式，20个类别，checkpoint使用 > >  > > ./tools/dist_train.sh /workspace/Co-DETR/projects/configs/co_dino_vit/co_dino_5scale_vit_large_coco_instance.py 4 /workspace/train_workdir 训练100 epoch，loss_mask: 20.5192，推理结果完全不对，除了数据集相关类别标签、数量之外，还有哪里需要修改吗？ 你好，我能先问问你修改了什么才能进行实例分割么

Thank you very much for sharing, I encountered the same problem, my situation is that it can be run with windows, but with docker under windows, this problem occurs, in...

> ./tools/dist_test.sh \ > configs/mask_rcnn_r50_fpn_1x_coco.py \ > checkpoints/mask_rcnn_r50_fpn_1x_coco_20200205-d4b0c5d6.pth \ > 8 \ > --eval bbox segm > > 把这个命令里的config和checkpoint换成对应instance seg的就可以进行测试 你好，如果我想使用co-detr做实例分割的训练，也是这么更改么，因为我没有发现readme中有提出相关的示范

>  > > in the file :gradient_checkpointing.py, I changed the code like in the image, then it works, the data was transfered to gpu instead of cpu. 感谢分享，我遇到了同样的问题，我看有个issue 里有老哥说的很明白：https://github.com/unslothai/unsloth/issues/1744#issuecomment-2772055277