解决Yolov8训练train时box_loss cls_loss dfl_loss为nan和 Box(P R mAP50 mAP50-95)为0问题和wandb.errors.UsageError

Ultralytics YOLOv8.0.202

 Python-3.9.16

torch-1.13.1

CUDA:11.6 (NVIDIA GeForce GTX 1650, 4096MiB)

Yolov8在训练时box_loss cls_loss dfl_loss为nan和 Box(P R mAP50 mAP50-95)为0的问题

第一步：找到ultralytics\cfg\default.yaml中的default.yaml文件

将default.yaml中的amp设置为false

amp: False # (bool) Automatic Mixed Precision (AMP) training, choices=[True, False], True runs AMP check

将default.yaml中的half设置为false

half: False  # (bool) use half precision (FP16)

第二步：

找到ultralytics\engine\validator.py中的validator.py文件

将self.args.half = self.device.type != 'cpu' # force FP16 val during training这行代码注释掉

# self.args.half = self.device.type != 'cpu'  # force FP16 val during training

修改之后的代码为

self.training = trainer is not None
        augment = self.args.augment and (not self.training)
        if self.training:
            self.device = trainer.device
            self.data = trainer.data
            # self.args.half = self.device.type != 'cpu'  # force FP16 val during training
            model = trainer.ema.ema or trainer.model
            model = model.half() if self.args.half else model.float()
            if hasattr(model, 'criterion'):
                if hasattr(model.criterion.bbox_loss, 'wiou_loss'):
                    model.criterion.bbox_loss.wiou_loss.eval()
            # self.model = model
            self.loss = torch.zeros_like(trainer.loss_items, device=trainer.device)
            self.args.plots &= trainer.stopper.possible_stop or (trainer.epoch == trainer.epochs - 1)
            model.eval()

修改之后的运行结果不再为0和nan

训练时遇到的问题2：

wandb.errors.UsageError: Invalid project name ‘runs/train‘: cannot contain characters ‘/,\\,#,?,%,found ' / '

找到ultralytics\utils\callbacks\wb.py中的wb.py文件

将trainer.args.project改为os.path.basename(trainer.args.project)并import os:

修改之后的代码为：

def on_pretrain_routine_start(trainer):
    import os
    """Initiate and start project if module is present."""
    wb.run or wb.init(project=os.path.basename(trainer.args.project) or 'YOLOv8', name=trainer.args.name, config=vars(trainer.args))

训练遇到的问题3：

页面文件太小，无法完成操作。 Error loading "D:\Anaconda\anaconda3\envs\yolo\lib\site-packages\torch\lib\shm.dll" or one of its dependencies.

ultralytics-main\ultralytics\cfg\default.yaml

里面把batch调整为4

workers调整为0

又可以正常训练了，哦耶！

个人训练报错记录，等有错误解决之后再补充！！