pytorch下对标签target进行one-hot编码

pytorch 进行one-hot编码的两种方法

torch.nn.functional 中的ont_hot方法

target = torch.tensor([0,1,2,3,4])
target = F.one_hot(target,num_classes = 5)
tensor([[1, 0, 0, 0, 0],
        [0, 1, 0, 0, 0],
        [0, 0, 1, 0, 0],
        [0, 0, 0, 1, 0],
        [0, 0, 0, 0, 1]])

使用scatter_函数

target = torch.tensor([0,2,3,4,1])
index = target.view(-1,1).long()
target = torch.zeros(target.size(0),5,dtype=torch.float32)
target = target.scatter_(1,index,1)
tensor([[1., 0., 0., 0., 0.],
        [0., 0., 1., 0., 0.],
        [0., 0., 0., 1., 0.],
        [0., 0., 0., 0., 1.],
        [0., 1., 0., 0., 0.]])
  • 1
    点赞
  • 1
    收藏
    觉得还不错? 一键收藏
  • 1
    评论

“相关推荐”对你有帮助么?

  • 非常没帮助
  • 没帮助
  • 一般
  • 有帮助
  • 非常有帮助
提交
评论 1
添加红包

请填写红包祝福语或标题

红包个数最小为10个

红包金额最低5元

当前余额3.43前往充值 >
需支付:10.00
成就一亿技术人!
领取后你会自动成为博主和红包主的粉丝 规则
hope_wisdom
发出的红包
实付
使用余额支付
点击重新获取
扫码支付
钱包余额 0

抵扣说明:

1.余额是钱包充值的虚拟货币,按照1:1的比例进行支付金额的抵扣。
2.余额无法直接购买下载,可以购买VIP、付费专栏及课程。

余额充值