blocks|key|5904622|text|尝试torch.repeat_interleave()方法：https://pytorch.org/docs/stable/torch.html#torch.repeat_interleave|type|unstyled|depth|inlineStyleRanges|entityRanges|offset|length|data|5904623|>>>+x+=+torch.tensor([1,+2,+3])
>>>+x.repeat_interleave(2)
tensor([1,+1,+2,+2,+3,+3])|code-block|syntax|javascript|5904624|entityMap|0|LINK|mutability|MUTABLE|url|https://pytorch.org/docs/stable/torch.html#torch.repeat_interleave^0|U|1U|0|0|0^^$0|@$1|2|3|4|5|6|7|Q|8|@]|9|@$A|R|B|S|1|T]]|C|$]]|$1|D|3|E|5|F|7|U|8|@]|9|@]|C|$G|H]]|$1|I|3|-4|5|6|7|V|8|@]|9|@]|C|$]]]|J|$K|$5|L|M|N|C|$O|P]]]]

Try torch.repeat_interleave() method: <a href="https://pytorch.org/docs/stable/torch.html#torch.repeat_interleave" rel="noreferrer">https://pytorch.org/docs/stable/torch.html#torch.repeat_interleave</a>

<pre><code>&gt;&gt;&gt; x = torch.tensor([1, 2, 3])
&gt;&gt;&gt; x.repeat_interleave(2)
tensor([1, 1, 2, 2, 3, 3])
</code></pre>

blocks|key|2376434|text|引用https://discuss.pytorch.org/t/how-to-tile-a-tensor/13853+-|type|unstyled|depth|inlineStyleRanges|entityRanges|offset|length|data|2376435|2376436|+z=+torch.FloatTensor([1,2,3,4,5,6,7,8,9])+1+2+3+4+5+6+7+8+9+z.transpose(0,1).repeat(1,3).view(-1，3).transpose(0,1)+1+1+2+2+2+3+3+3+4+4+4+5+5+5+6+6+7+7+7+8+8+8+9+9|blockquote|2376437|2376438|2376439|这将给你一个直观的感觉，它是如何工作的。|2376440|entityMap|0|LINK|mutability|MUTABLE|url|https://discuss.pytorch.org/t/how-to-tile-a-tensor/13853^0|2|1K|0|0|0|0|0|0|0^^$0|@$1|2|3|4|5|6|7|T|8|@]|9|@$A|U|B|V|1|W]]|C|$]]|$1|D|3|-4|5|6|7|X|8|@]|9|@]|C|$]]|$1|E|3|F|5|G|7|Y|8|@]|9|@]|C|$]]|$1|H|3|-4|5|6|7|Z|8|@]|9|@]|C|$]]|$1|I|3|-4|5|6|7|10|8|@]|9|@]|C|$]]|$1|J|3|K|5|6|7|11|8|@]|9|@]|C|$]]|$1|L|3|-4|5|6|7|12|8|@]|9|@]|C|$]]]|M|$N|$5|O|P|Q|C|$R|S]]]]

Quoting <a href="https://discuss.pytorch.org/t/how-to-tile-a-tensor/13853" rel="noreferrer">https://discuss.pytorch.org/t/how-to-tile-a-tensor/13853</a> - 

<blockquote>
<pre><code>z = torch.FloatTensor([[1,2,3],[4,5,6],[7,8,9]])
1 2 3
4 5 6
7 8 9
z.transpose(0,1).repeat(1,3).view(-1, 3).transpose(0,1)
1 1 1 2 2 2 3 3 3
4 4 4 5 5 5 6 6 6
7 7 7 8 8 8 9 9 9
</code></pre>
</blockquote>

This will give you a intuitive feel of how it works.

blocks|key|2376364|text|a+=+torch.Tensor([1,2,3,4])|type|code-block|depth|inlineStyleRanges|entityRanges|data|syntax|javascript|2376365|为了得到[1.,+2.,+3.,+4.,+1.,+2.,+3.,+4.,+1.,+2.,+3.,+4.]，我们在第一维重复张量三次：|unstyled|offset|length|style|CODE|2376366|a.repeat(3)|2376367|为了得到[1,1,1,2,2,2,3,3,3,4,4,4]，我们向张量添加一个维度，并在第二个维度中重复三次，以获得4+x+3张量，我们可以将其展平。|2376368|b+=+a.reshape(4,1).repeat(1,3).flatten()|2376369|或|2376370|b+=+a.reshape(4,1).repeat(1,3).view(-1)|2376371|entityMap^0|0|4|1C|0|0|4|P|1M|5|0|0|0|0^^$0|@$1|2|3|4|5|6|7|W|8|@]|9|@]|A|$B|C]]|$1|D|3|E|5|F|7|X|8|@$G|Y|H|Z|I|J]]|9|@]|A|$]]|$1|K|3|L|5|6|7|10|8|@]|9|@]|A|$B|C]]|$1|M|3|N|5|F|7|11|8|@$G|12|H|13|I|J]|$G|14|H|15|I|J]]|9|@]|A|$]]|$1|O|3|P|5|6|7|16|8|@]|9|@]|A|$B|C]]|$1|Q|3|R|5|F|7|17|8|@]|9|@]|A|$]]|$1|S|3|T|5|6|7|18|8|@]|9|@]|A|$B|C]]|$1|U|3|-4|5|F|7|19|8|@]|9|@]|A|$]]]|V|$]]

<pre><code>a = torch.Tensor([1,2,3,4])
</code></pre>
To get <code>[1., 2., 3., 4., 1., 2., 3., 4., 1., 2., 3., 4.]</code> we repeat the tensor thrice in the 1st dimension:
<pre><code>a.repeat(3)
</code></pre>
To get <code>[1,1,1,2,2,2,3,3,3,4,4,4]</code> we add a dimension to the tensor and repeat it thrice in the 2nd dimension to get a <code>4 x 3</code> tensor, which we can flatten.
<pre><code>b = a.reshape(4,1).repeat(1,3).flatten()
</code></pre>
or
<pre><code>b = a.reshape(4,1).repeat(1,3).view(-1)
</code></pre>

blocks|key|1805526|text|这是一个在张量中重复元素的泛型函数。|type|unstyled|depth|inlineStyleRanges|entityRanges|data|1805527|def+repeat(tensor,+dims):
++++if+len(dims)+!=+len(tensor.shape):
++++++++raise+ValueError("The+length+of+the+second+argument+must+equal+the+number+of+dimensions+of+the+first.")
++++for+index,+dim+in+enumerate(dims):
++++++++repetition_vector+=+[1]*(len(dims)%2B1)
++++++++repetition_vector[index%2B1]+=+dim
++++++++new_tensor_shape+=+list(tensor.shape)
++++++++new_tensor_shape[index]+*=+dim
++++++++tensor+=+tensor.unsqueeze(index%2B1).repeat(repetition_vector).reshape(new_tensor_shape)
++++return+tensor|code-block|syntax|javascript|1805528|如果你有|1805529|foo+=+tensor([[1,+2],
++++++++++++++[3,+4]])|1805530|通过调用repeat(foo,+[2,1])，您将获得|offset|length|style|CODE|1805531|tensor([[1,+2],
++++++++[1,+2],
++++++++[3,+4],
++++++++[3,+4]])|1805532|因此，您复制了维度0上的每个元素和维度1上的左侧元素。|1805533|entityMap^0|0|0|0|0|4|I|0|0|0^^$0|@$1|2|3|4|5|6|7|W|8|@]|9|@]|A|$]]|$1|B|3|C|5|D|7|X|8|@]|9|@]|A|$E|F]]|$1|G|3|H|5|6|7|Y|8|@]|9|@]|A|$]]|$1|I|3|J|5|D|7|Z|8|@]|9|@]|A|$E|F]]|$1|K|3|L|5|6|7|10|8|@$M|11|N|12|O|P]]|9|@]|A|$]]|$1|Q|3|R|5|D|7|13|8|@]|9|@]|A|$E|F]]|$1|S|3|T|5|6|7|14|8|@]|9|@]|A|$]]|$1|U|3|-4|5|6|7|15|8|@]|9|@]|A|$]]]|V|$]]

Here's a generic function that repeats elements in tensors.

<pre class="lang-py prettyprint-override"><code>def repeat(tensor, dims):
 if len(dims) != len(tensor.shape):
 raise ValueError("The length of the second argument must equal the number of dimensions of the first.")
 for index, dim in enumerate(dims):
 repetition_vector = [1]*(len(dims)+1)
 repetition_vector[index+1] = dim
 new_tensor_shape = list(tensor.shape)
 new_tensor_shape[index] *= dim
 tensor = tensor.unsqueeze(index+1).repeat(repetition_vector).reshape(new_tensor_shape)
 return tensor
</code></pre>

If you have 

<pre><code>foo = tensor([[1, 2],
 [3, 4]])
</code></pre>

By calling <code>repeat(foo, [2,1])</code> you get

<pre><code>tensor([[1, 2],
 [1, 2],
 [3, 4],
 [3, 4]])
</code></pre>

So you duplicated every element along dimension 0 and left elements as they are on dimension 1.

blocks|key|1809203|text|使用einops：|type|unstyled|depth|inlineStyleRanges|entityRanges|data|1809204|from+einops+import+repeat

repeat(x,+'i+->+(repeat+i)',+repeat=3)
#+like+{1,2,3,4,1,2,3,4,1,2,3,4}

repeat(x,+'i+->+(i+repeat)',+repeat=3)
#+like+{1,1,1,2,2,2,3,3,3,4,4,4}|code-block|syntax|javascript|1809205|这段代码适用于任何框架(numpy、torch、tf等)。|1809206|entityMap^0|0|0|0^^$0|@$1|2|3|4|5|6|7|K|8|@]|9|@]|A|$]]|$1|B|3|C|5|D|7|L|8|@]|9|@]|A|$E|F]]|$1|G|3|H|5|6|7|M|8|@]|9|@]|A|$]]|$1|I|3|-4|5|6|7|N|8|@]|9|@]|A|$]]]|J|$]]

Use einops:
<pre><code>from einops import repeat

repeat(x, 'i -&gt; (repeat i)', repeat=3)
# like {1,2,3,4,1,2,3,4,1,2,3,4}

repeat(x, 'i -&gt; (i repeat)', repeat=3)
# like {1,1,1,2,2,2,3,3,3,4,4,4}
</code></pre>
This code works identically for any framework (numpy, torch, tf, etc.)

blocks|key|2376510|text|你能试试这样的东西吗：|type|unstyled|depth|inlineStyleRanges|offset|length|style|BOLD|entityRanges|data|2376511|import+torch+as+pt

#1+work+as+numpy+tile

b+=+pt.arange(10)
print(b.repeat(3))

#2+work+as+numpy+tile

b+=+pt.tensor(1).repeat(10).reshape(2,-1)
print(b)

#3+work+as+numpy+repeat

t+=+pt.tensor([1,2,3])
t.repeat(2).reshape(2,-1).transpose(1,0).reshape(-1)|code-block|syntax|javascript|2376512|entityMap^0|0|B|0|0^^$0|@$1|2|3|4|5|6|7|M|8|@$9|N|A|O|B|C]]|D|@]|E|$]]|$1|F|3|G|5|H|7|P|8|@]|D|@]|E|$I|J]]|$1|K|3|-4|5|6|7|Q|8|@]|D|@]|E|$]]]|L|$]]

Can you try something like:

<pre><code>import torch as pt

#1 work as numpy tile

b = pt.arange(10)
print(b.repeat(3))

#2 work as numpy tile

b = pt.tensor(1).repeat(10).reshape(2,-1)
print(b)

#3 work as numpy repeat

t = pt.tensor([1,2,3])
t.repeat(2).reshape(2,-1).transpose(1,0).reshape(-1)
</code></pre>

I am trying to repeat a tensor in torch in two ways. For example repeating the tensor <code>{1,2,3,4}</code> 3 times both ways to yield;

<pre><code>{1,2,3,4,1,2,3,4,1,2,3,4}
{1,1,1,2,2,2,3,3,3,4,4,4}
</code></pre>

There is a built in torch:repeatTensor function which will generate the first of the two (like <code>numpy.tile()</code>) but I can't find one for the latter (like <code>numpy.repeat()</code>). I'm sure that I could call sort on the first to give the second but I think this might be computationally expensive for larger arrays?

Thanks.

Torch - repeat tensor like numpy repeat

翻译质量差，导致语言生硬或混乱。

没有提供实际的解决方法或示例。

解答不清晰，无法理解或解决问题。

页面排版不美观，阅读体验差。

文章

问答

视频

学习中心

腾讯云实验室

直播

竞赛

腾讯云架构师技术同盟交流圈

腾讯云数据库专区

腾讯云顾问专区

腾讯云原生专区

腾讯混元专区

腾讯云TCE专区

腾讯云Lighthouse专区

腾讯云HAI专区

腾讯云Edgeone专区

腾讯云存储专区

腾讯云智能专区

腾讯轻联专区 

腾讯云开发专区

TAPD专区

腾讯轻量云游戏服专区

腾讯云最具价值专家

腾讯云架构师技术同盟

腾讯云创作之星

腾讯云开发者先锋 

腾讯云代码助手

CODING DevOps

Cloud Studio

SDK中心

API中心

命令行工具

我正在尝试以两种方式重复torch中的张量。例如，将张量{1,2,3,4}重复两种方式3次以产生结果；{1,2,3,4,1,2,3,4,1,2,3,4}{1,1,1,2,2,2,3,3,3,4,4,4}有一个内置的torch: generate张量函数，它将生成两个函数中的第一个(如numpy.tile())，但我找不到一个用于后者(如numpy.repeat())。我确信我可以在第一个数组上调用

问Torch - repeat张量类numpy repeat
EN

回答 6

Stack Overflow用户

Stack Overflow用户

Stack Overflow用户

社区

活动

圈层

关于

腾讯云开发者

热门产品

热门推荐

更多推荐

问Torch - repeat张量类numpy repeatEN

回答 6

Stack Overflow用户

Stack Overflow用户

Stack Overflow用户

社区

活动

圈层

关于

腾讯云开发者

热门产品

热门推荐

更多推荐

问Torch - repeat张量类numpy repeat
EN