ValueError: only one element tensors can be converted to Python scalars when using torch.Tensor on list of tensors

Question

I have a list of tensors:

object_ids = [tensor([2., 3.]), tensor([2., 3.]), tensor([2., 3.]), tensor([2., 3.]), tensor([2., 3.]), tensor([2., 3.]), tensor([2., 3.]), tensor([2., 3.]), tensor([2., 3.]), tensor([2., 3.])]

Intuitively, it seems like I should be able to create a new tensor from this:

torch.as_tensor(object_ids, dtype=torch.float32)

But this does NOT work. Apparently, torch.as_tensor and torch.Tensor can only turn lists of scalars into new tensors. it cannot turn a list of d-dim tensors into a d+1 dim tensor.

Answer 1

You can use torch.stack .

In your example:

>>> object_ids = [tensor([2., 3.]), tensor([2., 3.]), tensor([2., 3.]), tensor([2., 3.]), tensor([2., 3.]), tensor([2., 3.]), tensor([2., 3.]), tensor([2., 3.]), tensor([2., 3.]), tensor([2., 3.])]
>>> torch.stack(object_ids)
tensor([[2., 3.],
        [2., 3.],
        [2., 3.],
        [2., 3.],
        [2., 3.],
        [2., 3.],
        [2., 3.],
        [2., 3.],
        [2., 3.],
        [2., 3.]])