torch.take# torch.take(input, index) → Tensor# Returns a new tensor with the elements of input at the given indices. The input tensor is treated as if it were viewed as a 1-D tensor. The result takes the same shape as the indices. Parameters input (Tensor) – the input tensor. index (LongTensor) – the indices into tensor Example: >>> src = torch.tensor([[4, 3, 5], ... [6, 7, 8]]) >>> torch.take(src, torch.tensor([0, 2, 5])) tensor([ 4, 5, 8])