learning problems

• TypeError: can't convert cuda:0 device type tensor to numpy. Use Tensor.cpu() to copy the tensor to host memory first.

Traceback (most recent call last):
  File "train.py", line 436, in <module>
    train(hyp)  # train normally
  File "train.py", line 330, in train
    results, maps = test.test(cfg,
  File "/home/roboticsai/anzhiqiang/yolov3/test.py", line 173, in test
    plot_images(imgs, output_to_target(output, width, height), paths=paths, names=names, fname=f)  # predictions
  File "/home/roboticsai/anzhiqiang/yolov3/utils/utils.py", line 821, in output_to_target
    return np.array(targets)
  File "/home/roboticsai/.virtualenvs/azq/lib/python3.8/site-packages/torch/tensor.py", line 642, in __array__
    return self.numpy()
TypeError: can't convert cuda:0 device type tensor to numpy. Use Tensor.cpu() to copy the tensor to host memory first.

解决方法：

打开/home/roboticsai/.virtualenvs/azq/lib/python3.8/site-packages/torch/tensor.py文件。在两个return self后边加上cpu()。

• KeyError: 'lrf'   

解决方法： 

根据'lrf'所在代码

 lf = lambda x: ((1 + math.cos(x * math.pi / epochs)) / 2) * (1 - hyp['lrf']) + hyp['lrf']

找到调用该参数的函数，即hyp ，找到下面这段代码

parser.add_argument('--hyp', type=str, default='data/hyp.scratch.yaml', help='hyperparameters path')

然后再找到data/hyp.scratch.yaml文件

把'lrf'函数按照文件内的格式加进去。