pytorch拼接与拆分

**

一、拼接

**
cat/stack
cat在指定的维度上进行连接;
stack创建了新的维度进行连接。

In [1]: import torch

In [2]: a = torch.rand(4,32,8)

In [3]: b = torch.rand(5,32,8)

In [4]: torch.cat([a,b],dim=0).shape
Out[4]: torch.Size([9, 32, 8])

In [5]: a1 = torch.rand(4,3,32,32)

In [6]: a2 = torch.rand(5,3,32,32)

In [7]: torch.cat([a1,a2],dim=0).shape
Out[7]: torch.Size([9, 3, 32, 32])

In [8]: a2 = torch.rand(4,1,32,32)

In [9]: torch.cat([a1,a2],dim=1).shape
Out[9]: torch.Size([4, 4, 32, 32])

In [10]: a1 = torch.rand(4,3,16,32)

In [11]: a2 = torch.rand(4,3,16,32)

In [12]: torch.cat([a1,a2],dim=2).shape
Out[12]: torch.Size([4, 3, 32, 32])
In [10]: a1 = torch.rand(4,3,16,32)

In [11]: a2 = torch.rand(4,3,16,32)

In [12]: torch.cat([a1,a2],dim=2).shape
Out[12]: torch.Size([4, 3, 32, 32])

In [13]: torch.stack([a1,a2],dim=2).shape
Out[13]: torch.Size([4, 3, 2, 16, 32])

In [14]: a = torch.rand(32,8)

In [15]: b = torch.rand(32,8)

In [16]: torch.stack([a,b],dim=0).shape
Out[16]: torch.Size([2, 32, 8])

**

二、拆分

**
split/chunk
split根据每个单元固定长度来拆分或者根据需要的长度来拆分
chunk按照数量来拆分,给定的参数为需要拆分的数量

In [14]: a = torch.rand(32,8)

In [15]: b = torch.rand(32,8)

In [16]: torch.stack([a,b],dim=0).shape
Out[16]: torch.Size([2, 32, 8])

In [17]: a = torch.rand(32,8)

In [18]: b = torch.rand(32,8)

In [19]: c = torch.stack([a,b],dim=0)

In [20]: a.shape,b.shape,c.shape
Out[20]: (torch.Size([32, 8]), torch.Size([32, 8]), torch.Size([2, 32, 8]))

In [21]: aa,bb = c.split([1,1],dim=0)

In [22]: aa.shape,bb.shape
Out[22]: (torch.Size([1, 32, 8]), torch.Size([1, 32, 8]))

In [23]: aa,bb = c.split(1,dim=0)

In [24]: aa.shape,bb.shape
Out[24]: (torch.Size([1, 32, 8]), torch.Size([1, 32, 8]))

In [25]: aa,bb = c.split(2,dim=0)
---------------------------------------------------------------------------
ValueError                                Traceback (most recent call last)
<ipython-input-25-27b6f4946a79> in <module>
----> 1 aa,bb = c.split(2,dim=0)

ValueError: not enough values to unpack (expected 2, got 1)
In [32]: a = torch.rand(6,32,8)

In [33]: aa,bb,cc = a.split(2,dim=0)#固定长度拆分

In [34]: aa.shape,bb.shape,cc.shape
Out[34]: (torch.Size([2, 32, 8]), torch.Size([2, 32, 8]), torch.Size([2, 32, 8]))

In [35]: aa,bb,cc = a.split([1,3,2],dim=0)#按照实际需求长度拆分

In [36]: aa.shape,bb.shape,cc.shape
Out[36]: (torch.Size([1, 32, 8]), torch.Size([3, 32, 8]), torch.Size([2, 32, 8]))
In [26]: aa,bb = c.chunk(2,dim=0)

In [27]: aa.shape,bb.shape
Out[27]: (torch.Size([1, 32, 8]), torch.Size([1, 32, 8]))

In [37]: aa,bb,cc = a.chunk(3,dim=0)

In [38]: aa.shape,bb.shape,cc.shape
Out[38]: (torch.Size([2, 32, 8]), torch.Size([2, 32, 8]), torch.Size([2, 32, 8]))
  • 0
    点赞
  • 1
    收藏
    觉得还不错? 一键收藏
  • 0
    评论
评论
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