Pytorch训练SSD时出现的警告和错误（更新中）

1.  ERROR : assert win is not None, 'Must define a window to update' AssertionError: Must define a wind（train.py中修改)

上图是修改完后的样子，将这段函数上面代码中的epoch+=1挪到 “if args.visdom:" 函数下，并把update_vis_plot括号中的内容替换成图片中的内容，问题解决，具体可参考https://github.com/amdegroot/ssd.pytorch/issues/234

2.   VisibleDeprecationWarning: Creating an ndarray from ragged nested sequences (which is a list-or-tuple of lists-or-tuples-or ndarrays with different lengths or shapes) is deprecated. If you meant to do this, you must specify 'dtype=object' when creating the ndarray.

/home/bb622/ssd.pytorch-master/utils/augmentations.py:239: VisibleDeprecationWarning: Creating an ndarray from ragged nested sequences (which is a list-or-tuple of lists-or-tuples-or ndarrays with different lengths or shapes) is deprecated. If you meant to do this, you must specify 'dtype=object' when creating the ndarray.
  mode = random.choice(self.sample_options)

出现这种错误要么降低numpy版本，降到19.0以下，要么添加dtypt = object（尝试了几次，没有成功，可能添加dtype的地方不对）

3.   UserWarning: volatile was removed and now has no effect. Use `with torch.no_grad():` instead.（train.py中修改）

修改前

        if args.cuda:
            images = Variable(images.cuda())
            targets = [Variable(ann.cuda(), volatile=True) for ann in targets]
        else:
            images = Variable(images)
            targets = [Variable(ann, volatile=True) for ann in targets]

修改后

        if args.cuda:
            images = Variable(images.cuda())
            with torch.no_grad():
                targets = [Variable(ann.cuda()) for ann in targets]
        else:
            images = Variable(images)
            with torch.no_grad():
                targets = [Variable(ann) for ann in targets]

4.  UserWarning: size_average and reduce args will be deprecated, please use reduction='sum' instead.
 

 UserWarning: size_average and reduce args will be deprecated, please use reduction='sum' instead.
  warnings.warn(warning.format(ret))

将 multibox_loss.py函数中的90行和111行的

loss_l = F.smooth_l1_loss(loc_p, loc_t, size_average=False)改为

loss_l = F.smooth_l1_loss(loc_p, loc_t, reduction='sum')