dmme#

LightningModules#

DataModules#

Common functions#

`gaussian`	Samples from gaussian with specified shape, dtype, device using torch.randn
`gaussian_like`	Samples from gaussian like the tensor x using torch.randn_like
`uniform_int`	Samples ints from uniform distribution using torch.randint
`norm`	Normalize input to \([-1, 1]\) linearly
`denorm`	Denormalize input normalized to \([-1, 1]\) linearly back to \([0, 1]\)
`make_history`	Visualize diffusion process given an array of histories

dmme.gaussian(shape, dtype=None, device=None)[source]#: Samples from gaussian with specified shape, dtype, device using torch.randn

dmme.gaussian_like(x)[source]#: Samples from gaussian like the tensor x using torch.randn_like

dmme.uniform_int(min, max, count=1, device=None)[source]#: Samples ints from uniform distribution using torch.randint

dmme.norm(x)[source]#: Normalize input to \([-1, 1]\) linearly

dmme.denorm(x)[source]#: Denormalize input normalized to \([-1, 1]\) linearly back to \([0, 1]\)

dmme.make_history(history)[source]#

Visualize diffusion process given an array of histories

Parameters:: history (List[torch.Tensor]) – list of diffusion history with each item as a tensor of shape \((N, C, H, W)\)