Tsardoz

Thanks but I am still confused. If I download any segformer model from their git even the B3 used as demo (and shows it working well) I get terrible results....

Thanks. I had seen that but the weights had not registered with me. It also says which backbone to use. On Sat, Jun 11, 2022 at 11:24 PM sithu3 ***@***.***>...

Thanks but that did not work either. I embedded the codes into class_names.py and it worked [class_names.txt](https://github.com/czczup/ViT-Adapter/files/9389584/class_names.txt) .

Thanks. Do I still need to define the model if it is loaded by load_from? If I delete this I get the error _KeyError: 'XCiT is not in the models...

Thanks. I tried using the decorator method first as I had already written the training code. This did not work. When I looked at train_from_recipe I saw that it needed...

Of course the comment even says # YOLO**v5** Albumentations class (optional, only used if package is installed). It would help if I read stuff! There is nothing in the path...

Sorry @Laughing-q I still cannot get this to work. With this command for training !yolo task=detect mode=train model=yolov8s.pt data=fallset.yaml device=\'0,1,2,3,4,5\' epochs=20 batch=480 lr0=0.001 save=True and this in the augment.py file...

@Laughing-q that is very strange. Are these lines 549, 550 and 551?

@Laughing-q OK I have reinstalled and tried again. Still nothing. I added this last line and nothing is printed 

I will try debugging this