Unflatten#

classtorch.nn.modules.flatten.Unflatten(dim,unflattened_size)[source]#

Unflattens a tensor dim expanding it to a desired shape. For use withSequential.

• dim specifies the dimension of the input tensor to be unflattened, and it canbe eitherint orstr whenTensor orNamedTensor is used, respectively.

• unflattened_size is the new shape of the unflattened dimension of the tensor and it can beatuple of ints or alist of ints ortorch.Size forTensor input; aNamedShape(tuple of(name, size) tuples) forNamedTensor input.

Shape:

• Input:$(\ast ,S_{\text{dim}},\ast )$(∗,Sdim​,∗), where$S_{\text{dim}}$Sdim​ is the size atdimensiondim and$\ast$∗ means any number of dimensions including none.

• Output:$(\ast ,U_1,\mathrm{.}\mathrm{.}\mathrm{.},U_n,\ast )$(∗,U1​,...,Un​,∗), where$U$U =unflattened_size and${\prod }_{i=1}^n{U}_i=S_{\text{dim}}$∏i=1n​Ui​=Sdim​.

Parameters

• dim (Union[int,str]) – Dimension to be unflattened

• unflattened_size (Union[torch.Size,Tuple,List,NamedShape]) – New shape of the unflattened dimension

Examples

>>>input=torch.randn(2,50)>>># With tuple of ints>>>m=nn.Sequential(>>>nn.Linear(50,50),>>>nn.Unflatten(1,(2,5,5))>>>)>>>output=m(input)>>>output.size()torch.Size([2, 2, 5, 5])>>># With torch.Size>>>m=nn.Sequential(>>>nn.Linear(50,50),>>>nn.Unflatten(1,torch.Size([2,5,5]))>>>)>>>output=m(input)>>>output.size()torch.Size([2, 2, 5, 5])>>># With namedshape (tuple of tuples)>>>input=torch.randn(2,50,names=("N","features"))>>>unflatten=nn.Unflatten("features",(("C",2),("H",5),("W",5)))>>>output=unflatten(input)>>>output.size()torch.Size([2, 2, 5, 5])

NamedShape#

alias oftuple[tuple[str,int]]

extra_repr()[source]#

Returns the extra representation of the module.

Return type

str

forward(input)[source]#

Runs the forward pass.

Return type

Tensor