總結：按元素相乘用torch.mul，二維矩陣乘法用torch.mm，batch二維矩陣用torch.bmm，batch、廣播用torch.matmul
if __name__==
"__main__"
: a = torch.tensor([1
,2,3
])b = torch.arange(0,
12).reshape((4
,3))
c = torch.tensor([4
,5,6
,7])
d = torch.arange(0,
12).reshape((3
,4))
aa = torch.unsqueeze(a, dim=1)
cc = torch.unsqueeze(c, dim=0)
print
('a:,a.shape:'
.format
(a, a.shape)
)print
('b:,b.shape:'
.format
(b, b.shape)
)print
('c:,c.shape:'
.format
(c, c.shape)
)print
('d:,d.shape:'
.format
(d, d.shape)
)print
('aa:,aa.shape:'
.format
(aa, aa.shape)
)print
('cc:,cc.shape:'
.format
(cc, cc.shape)
)#output:
a:tensor([1
,2,3
]),a.shape:torch.size([3
])b:tensor([[
0,1,
2],[
3,4,
5],[
6,7,
8],[
9,10,
11]])
,b.shape:torch.size([4
,3])
c:tensor([4
,5,6
,7])
,c.shape:torch.size([4
])d:tensor([[
0,1,
2,3]
,[4,
5,6,
7],[
8,9,
10,11]
]),d.shape:torch.size([3
,4])
aa:tensor([[
1],[
2],[
3]])
,aa.shape:torch.size([3
,1])
cc:tensor([[
4,5,
6,7]
]),cc.shape:torch.size([1
,4])
print
(a.mul(b)
)#a.shape:torch.size([3])
print
(aa.mul(d)
)#aa.shape:torch.size([3, 1])
#output:
tensor([[
0,2,
6],[
3,8,
15],[
6,14,
24],[
9,20,
33]])
tensor([[
0,1,
2,3]
,[8,
10,12,
14],[
24,27,
30,33]
])
print
(b.mm(d)
)#output：
tensor([[
20,23,
26,29]
,[56,
68,80,
92],[
92,113,
134,
155],[
128,
158,
188,
218]
])
情況1：向量✖向量：點乘
>>
>
# vector x vector
>>
> tensor1 = torch.randn(3)
>>
> tensor2 = torch.randn(3)
>>
> torch.matmul(tensor1, tensor2)
.size(
)torch.size(
)
如 tensor1 @ tensor2 --> (3,4)(4,) --> (3,4)(4,1) --> (3,1) --> (3,)
>>
>
# matrix x vector
>>
> tensor1 = torch.randn(3,
4)>>
> tensor2 = torch.randn(4)
>>
> torch.matmul(tensor1, tensor2)
.size(
)torch.size([3
])
如： tensor1 @ tensor2 --> (10，3，4)(4,) --> (10，3，4)(10，4，1) --> (10，3，1) --> (10，3)
>>
>
# batched matrix x broadcasted vector
>>
> tensor1 = torch.randn(10,
3,4)
>>
> tensor2 = torch.randn(4)
>>
> torch.matmul(tensor1, tensor2)
.size(
)torch.size([10
,3])
>>
>
# batched matrix x batched matrix
>>
> tensor1 = torch.randn(10,
3,4)
>>
> tensor2 = torch.randn(10,
4,5)
>>
> torch.matmul(tensor1, tensor2)
.size(
)torch.size([10
,3,5
])
>>
>
# batched matrix x broadcasted matrix
>>
> tensor1 = torch.randn(10,
3,4)
>>
> tensor2 = torch.randn(4,
5)>>
> torch.matmul(tensor1, tensor2)
.size(
)torch.size([10
,3,5
])
PyTorch 矩陣乘法總結
torch.mm mat1,mat2,out none 其中mat1 n times m mat2 m times d 輸出out的維度是 n times d 該函式一般只用來計算兩個二維矩陣的矩陣乘法，並且不支援broadcast操作。由於神經網路訓練一般採用mini batch，經常輸入的時三維...
Pytorch 中 torchvision的錯誤
在學習pytorch的時候，使用 torchvision的時候發生了乙個小小的問題 安裝都成功了，並且import torch也沒問題，但是在import torchvision的時候，出現了如下所示的錯誤資訊 dll load failed 找不到指定模組。首先，我們得知道torchvision在...
pytorch中index select 的用法
a torch.linspace 1,12,steps 12 view 3,4 print a b torch.index select a,0,torch.tensor 0,2 print b print a.index select 0,torch.tensor 0,2 c torch.inde...