RuntimeError: result type Float can't be cast to the desired output type long int

[poojatambe]

Facing error while training.
Training command: !python polygon_train.py --weights yolov5s.pt --cfg polygon_yolov5s_ucas.yaml
--data data/custom.yaml --hyp hyp.ucas.yaml --img-size 1024
--epochs 3 --batch-size 12 --noautoanchor --polygon --cache

[SjoerdBoonen]

Facing error while training. Training command: !python polygon_train.py --weights yolov5s.pt --cfg polygon_yolov5s_ucas.yaml --data data/custom.yaml --hyp hyp.ucas.yaml --img-size 1024 --epochs 3 --batch-size 12 --noautoanchor --polygon --cache

Hey Im currently getting the same error, did you find a way to fix it?
I looked at the original YOLOv5 repo from Utralytics and there they mentioned that they fixed the issue in a more recent version (ultralytics/yolov5#8405), however I think this repo is built on an older version.

@XinzeLee do you know how to fix this?

[dt140120]

Facing error while training. Training command: !python polygon_train.py --weights yolov5s.pt --cfg polygon_yolov5s_ucas.yaml --data data/custom.yaml --hyp hyp.ucas.yaml --img-size 1024 --epochs 3 --batch-size 12 --noautoanchor --polygon --cache

same error

[github-rajs]

Getting same error.Any fix for this?

[jlclemon]

I have found that modifying the offending line in loss.py to be:

indices.append((b, a, gj.clamp_(0, int(gain[3].item()) - 1), gi.clamp_(0, int(gain[2].item()) - 1))) # image, anchor, grid indices

allows it to run training. I am still verifying if the training results in a correct model.

[dt140120]

I have found that modifying the offending line in loss.py to be:

indices.append((b, a, gj.clamp_(0, int(gain[3].item()) - 1), gi.clamp_(0, int(gain[2].item()) - 1))) # image, anchor, grid indices

allows it to run training. I am still verifying if the training results in a correct model.

Can you capture the screen of training for reference ?

[helianthus1]

In loss.py (line 173 ), gain = torch.ones(7, device=targets.device) → gain = torch.ones(7, device=targets.device).long()

[poojatambe]

Run this command before python setup.py install :

!pip3 install torch==1.11.0+cu113 torchvision==0.12.0+cu113 torchaudio==0.11.0+cu113 -f https://download.pytorch.org/whl/cu113/torch_stable.html