样本加权损失

早前数据集比较少，使用了一些图像的权重增强策略

现在图像多了,  这些策略不需要了：

记录如下：

   sampler_weights =[]
   for indexx in range(len(train_dataset)):
       for aa,bb in dic_w.items():
            aaaa=str(train_dataset[indexx][1])
        #    print(aaaa)
            if aaaa in aa:
                ccc =float(bb)
                sampler_weights.append(ccc)
             #   print(sampler_weights)

    print(sampler_weights)
    train_nums = 2500
   train_sampler = torch.utils.data.WeightedRandomSampler(weights=sampler_weights, num_samples=train_nums, replacement=True)
    train_loader = torch.utils.data.DataLoader(
        train_dataset, batch_size=args.batch_size, shuffle=False,
        num_workers=args.workers, pin_memory=True, sampler=train_sampler)

因为使用了weightedrangomsampler 所以shuffle 要设置为false