DDMixin

class DDMixin(*args, **kwargs)

Bases: DDProtocol

Diffusion data mixin for common functionality.

clone() → Self

detach() → Self

Return a new instance detached from the current computation graph.

property device: device

gradient(outputs: Tensor, create_graph: bool = False, retain_graph: bool = False) → Self

Compute the gradient of output w.r.t. self.

Returns: An instance of Self containing the gradients.

ones_like() → Self

randn_like() → Self

requires_grad(mode: bool = True) → Self

In-place operation to start recording operations for autograd.

to(device: device | str) → Self

zeros_like() → Self