报错(interrupted by signal 11: SIGSEGV

最新推荐文章于 2024-07-01 18:10:00 发布

AADELAIDE

最新推荐文章于 2024-07-01 18:10:00 发布

阅读量3.5k

点赞数 1

分类专栏：类别不平衡

本文链接：https://blog.csdn.net/AADELAIDE/article/details/118676057

版权

类别不平衡专栏收录该内容

1 篇文章 0 订阅

订阅专栏

(interrupted by signal 11: SIGSEGV
准备做的是数据不平衡的问题，使用的是pytorch 交叉熵函数，weight参数一维tensor。
未注意到test的交叉熵损失函数不需要.cuda()形式，所以之跑出一个train就报错了
改为： weight_three = torch.FloatTensor([0.362, 4.752, 41.809])
criterion = torch.nn.CrossEntropyLoss(weight=weight_three.cuda())
criterion1 = torch.nn.CrossEntropyLoss()
测试集不需要类别不平衡惩罚啊

train_data.data_idx = get_train_idx(class1_train, class2_train, class3_train)
trainLoader = DataLoader(train_data, batch_size=args.batch_size, shuffle=True, num_workers=8)

    train_loss, train_f1 = train(net, trainLoader, criterion, optimizer, epoch, args)
    test_loss, test_f1, f1_0, f1_1, f1_2 = test(net, criterion1, testLoader, args, result)