nami.processes.diffusion

Classes

Diffusion(model, schedule, solver, *, ...[, ...])
DiffusionProcess(model, schedule, solver, *, ...)

class nami.processes.diffusion.Diffusion(model, schedule, solver, *, parameterization, t0=1.0, t1=0.0, base=None, event_shape=None, validate_args=True)

Bases: LazyDistribution

Parameters:

• parameterization (str)

• t0 (float)

• t1 (float)

• base (LazyDistribution | torch.distributions.Distribution | None)

• event_shape (tuple[int, ...] | None)

• validate_args (bool)

forward(c=None)

Define the computation performed at every call.

Should be overridden by all subclasses.

Note

Although the recipe for forward pass needs to be defined within this function, one should call the Module instance afterwards instead of this since the former takes care of running the registered hooks while the latter silently ignores them.

Return type:

DiffusionProcess

Parameters:

c (Tensor | None)

class nami.processes.diffusion.DiffusionProcess(model, schedule, solver, *, parameterization, t0=1.0, t1=0.0, base, base_scale=None, context=None, validate_args=True)

Bases: object

Parameters:

• parameterization (str)

• t0 (float)

• t1 (float)

• base (torch.distributions.Distribution)

• base_scale (torch.Tensor | None)

• context (torch.Tensor | None)

• validate_args (bool)

property batch_shape: tuple[int, ...]

property event_shape: tuple[int, ...]

rsample(sample_shape=(), *, guidance_fn=None)

Return type:

Tensor

sample(sample_shape=(), *, guidance_fn=None)

Return type:

Tensor