lucid.unbind¶
The unbind function removes a specified axis by slicing the tensor into multiple sub-tensors along that axis. It is functionally equivalent to splitting and squeezing the axis.
Function Signature¶
def unbind(a: Tensor, /, axis: int = 0) -> tuple[Tensor, ...]
Parameters¶
a (Tensor): The input tensor to be unbound.
axis (int, optional): The axis along which to unbind. Default is 0.
Returns¶
tuple[Tensor, …]: A tuple of sub-tensors where each tensor corresponds to one slice along the specified axis. Each output tensor will have one fewer dimension than the input tensor.
Tip
This operation is similar to split followed by squeeze on the target axis.
Backward Gradient Calculation¶
For input tensor \(\mathbf{a}\) unbound along an axis, and each output \(\mathbf{y}_i\), the gradient is propagated by embedding \(\nabla \mathbf{y}_i\) back into its slice:
Examples¶
>>> import lucid
>>> x = lucid.tensor([[1, 2], [3, 4], [5, 6]], requires_grad=True)
>>> y0, y1, y2 = lucid._util.unbind(x, axis=0)
>>> print(y0)
Tensor([1, 2], grad=None)
>>> y0.backward()
>>> print(x.grad)
[[1. 1.]
[0. 0.]
[0. 0.]]
Warning
The returned tensors are views and not copies. Modifying them directly may interfere with gradient propagation.