stack¶

dragon.vm.torch.stack(
  tensors,
  dim=0,
  out=None
)[source]¶

Stack the inputs along the given dimension.

All the dimensions of inputs should be same:

x1 = torch.ones(2, 3)
x2 = torch.zeros(2, 4)
y = torch.stack([x1, x1])  # Ok
z = torch.stack([x1, x2])  # Wrong

dim can be negative:

x = torch.tensor([[1, 2], [3, 4]])
y = torch.stack([x, x], dim=1)
z = torch.stack([x, x], dim=-1)  # Equivalent

Parameters:

• tensors (Sequence[dragon.vm.torch.Tensor]) – The input tensors.
• dim (int, optional, default=0) – The dimension to stack.
• out (dragon.vm.torch.Tensor, optional) – The output tensor.

Returns:

dragon.vm.torch.Tensor – The output tensor.