Python 如何在 PyTorch 中做矩阵的乘积
声明：本页面是StackOverFlow热门问题的中英对照翻译，遵循CC BY-SA 4.0协议，如果您需要使用它，必须同样遵循CC BY-SA许可，注明原文地址和作者信息，同时你必须将它归于原作者（不是我）：StackOverFlow
原文地址: http://stackoverflow.com/questions/44524901/
Warning: these are provided under cc-by-sa 4.0 license. You are free to use/share it, But you must attribute it to the original authors (not me):
StackOverFlow
How to do product of matrices in PyTorch
提问by blckbird
In numpy I can do a simple matrix multiplication like this:
在 numpy 中，我可以做一个简单的矩阵乘法，如下所示：
a = numpy.arange(2*3).reshape(3,2)
b = numpy.arange(2).reshape(2,1)
print(a)
print(b)
print(a.dot(b))
However, when I am trying this with PyTorch Tensors, this does not work:
但是，当我使用 PyTorch Tensors 尝试此操作时，这不起作用：
a = torch.Tensor([[1, 2, 3], [1, 2, 3]]).view(-1, 2)
b = torch.Tensor([[2, 1]]).view(2, -1)
print(a)
print(a.size())
print(b)
print(b.size())
print(torch.dot(a, b))
This code throws the following error:
此代码引发以下错误：
RuntimeError: inconsistent tensor size at /Users/soumith/code/builder/wheel/pytorch-src/torch/lib/TH/generic/THTensorMath.c:503
运行时错误：/Users/soumith/code/builder/wheel/pytorch-src/torch/lib/TH/generic/THTensorMath.c:503 处的张量大小不一致
Any ideas how matrix multiplication can be conducted in PyTorch?
任何想法如何在 PyTorch 中进行矩阵乘法？
回答by mexmex
You're looking for
您正在寻找
torch.mm(a,b)
Note that torch.dot()behaves differently to np.dot(). There's been some discussion about what would be desirable here. Specifically, torch.dot()treats both aand bas 1D vectors (irrespective of their original shape) and computes their inner product. The error is thrown, because this behaviour makes your aa vector of length 6 and your ba vector of length 2; hence their inner product can't be computed. For matrix multiplication in PyTorch, use torch.mm(). Numpy's np.dot()in contrast is more flexible; it computes the inner product for 1D arrays and performs matrix multiplication for 2D arrays.
请注意， 的torch.dot()行为与 不同np.dot()。有一些关于什么是可取的在这里讨论。具体来说，torch.dot()将a和 都b视为一维向量（不管它们的原始形状如何）并计算它们的内积。抛出错误，因为这种行为使您a成为长度为 6b的向量和长度为 2 的向量；因此无法计算它们的内积。对于 PyTorch 中的矩阵乘法，使用torch.mm(). np.dot()相比之下，Numpy更灵活；它计算一维数组的内积并为二维数组执行矩阵乘法。
By popular demand, the function torch.matmulperforms matrix multiplications if both arguments are 2Dand computes their dot product if both arguments are 1D. For inputs of such dimensions, its behaviour is the same as np.dot. It also lets you do broadcasting or matrix x matrix, matrix x vectorand vector x vectoroperations in batches. For more info, see its docs.
根据普遍的需求，torch.matmul如果两个参数都是，则该函数执行矩阵乘法，如果两个参数都是2D，则计算它们的点积1D。对于此类维度的输入，其行为与 相同np.dot。它还允许您批量进行广播或matrix x matrix,matrix x vector和vector x vector操作。有关更多信息，请参阅其文档。
# 1D inputs, same as torch.dot
a = torch.rand(n)
b = torch.rand(n)
torch.matmul(a, b) # torch.Size([])
# 2D inputs, same as torch.mm
a = torch.rand(m, k)
b = torch.rand(k, j)
torch.matmul(a, b) # torch.Size([m, j])
回答by BiBi
If you want to do a matrix (rank 2 tensor) multiplication you can do it in four equivalent ways:
如果你想做一个矩阵（秩 2 张量）乘法，你可以用四种等效的方式来做：
AB = A.mm(B) # computes A.B (matrix multiplication)
# or
AB = torch.mm(A, B)
# or
AB = torch.matmul(A, B)
# or, even simpler
AB = A @ B # Python 3.5+
There are a few subtleties. From the PyTorch documentation:
有一些微妙之处。从PyTorch 文档：
torch.mm does not broadcast. For broadcasting matrix products, see torch.matmul().
torch.mm 不广播。对于广播矩阵产品，请参见 torch.matmul()。
For instance, you cannot multiply two 1-dimensional vectors with torch.mm, nor multiply batched matrices (rank 3). To this end, you should use the more versatile torch.matmul. For an extensive list of the broadcasting behaviours of torch.matmul, see the documentation.
例如，您不能将两个一维向量与torch.mm相乘，也不能将批处理矩阵（等级 3）相乘。为此，您应该使用更通用的torch.matmul. 有关 的广播行为的详细列表torch.matmul，请参阅文档。
For element-wise multiplication, you can simply do (if A and B have the same shape)
对于逐元素乘法，您可以简单地执行（如果 A 和 B 具有相同的形状）
A * B # element-wise matrix multiplication (Hadamard product)
回答by David Jung
Use torch.mm(a, b)or torch.matmul(a, b)
Both are same.
用途torch.mm(a, b)或torch.matmul(a, b)
两者相同。
>>> torch.mm
<built-in method mm of type object at 0x11712a870>
>>> torch.matmul
<built-in method matmul of type object at 0x11712a870>
There's one more option that may be good to know.
That is @operator. @Simon H.
还有一个可能很高兴知道的选项。那就是@运营商。@西蒙H。
>>> a = torch.randn(2, 3)
>>> b = torch.randn(3, 4)
>>> a@b
tensor([[ 0.6176, -0.6743, 0.5989, -0.1390],
[ 0.8699, -0.3445, 1.4122, -0.5826]])
>>> a.mm(b)
tensor([[ 0.6176, -0.6743, 0.5989, -0.1390],
[ 0.8699, -0.3445, 1.4122, -0.5826]])
>>> a.matmul(b)
tensor([[ 0.6176, -0.6743, 0.5989, -0.1390],
[ 0.8699, -0.3445, 1.4122, -0.5826]])
The three give the same results.
三者给出相同的结果。
Related links:
Matrix multiplication operator
PEP 465 -- A dedicated infix operator for matrix multiplication
