pygsti.modelmembers.torchable

Defines the interface that ModelMembers must satisfy to be compatible with the PyTorch-backed forward simulator in pyGSTi/forwardsims/torchfwdsim.py.

Module Contents

Classes

Torchable

Base class for Model member objects that possess a definite state space, parameters count, and evolution type.

class pygsti.modelmembers.torchable.Torchable(state_space, evotype, gpindices=None, parent=None)

Bases: pygsti.modelmembers.modelmember.ModelMember

Base class for Model member objects that possess a definite state space, parameters count, and evolution type.

A ModelMember can be vectorized into/onto a portion of their parent Model’s (or other ModelMember’s) parameter vector. They therefore contain a gpindices reference to the global Model indices “owned” by this member. Note that GateSetMembers may contain other GateSetMembers (may be nested).

Parameters

state_spaceStateSpace

The state space, which should match the parent model if/when one exists.

evotypeEvoType or str

The evolution type, which should match the parent model if/when one exists.

gpindicesslice or numpy.ndarray, optional

The indices of this member’s local parameters into the parent Model’s parameter vector.

parentModel, optional

The parent model.

Attributes

dirtybool

Whether this member’s local parameters may have been updated without its parent’s knowledge. The parent model can check this flag and perform re-synchronization of it’s parameter vector when needed.

gpindicesslice or numpy.ndarray

The indices of this member’s local parameters into the parent Model’s parameter vector.

parentModel, optional

The parent model.

Initialize a new ModelMember

abstractmethod stateless_data() → Tuple

Return this ModelMember’s data that is considered constant for purposes of model fitting.

Note: the word “stateless” here is used in the sense of object-oriented programming.

abstractmethod static torch_base(sd: Tuple, t_param: torch.Tensor) → torch.Tensor

Suppose “obj” is an instance of some Torchable subclass. If we compute

vec = obj.to_vector() t_param = torch.from_numpy(vec) sd = obj.stateless_data() t = type(obj).torch_base(sd, t_param)

then t will be a PyTorch Tensor that represents “obj” in a canonical numerical way.

The meaning of “canonical” is implementation dependent. If type(obj) implements the .base attribute, then a reasonable implementation will probably satisfy

np.allclose(obj.base, t.numpy()).