adpeps.ipeps.ctm¶

Main CTM code

The individual site and boundary tensors come in a special list-type object (TList), which has extra indexing features, such as periodic boundary conditions and shift contexts

All ncon contractions are defined in contractions.yaml

class CTM(tensors, chi, conv_fun=None)[source]¶

CTM class

Parameters

tensors (adpeps.utils.ctmtensors.CTMTensors) – input ctmtensors
chi (int) – boundary bond dimension

show_progress()[source]¶: Print out the current progress

update()[source]¶: Perform an update of all boundary tensors

property converged¶: Check convergence with supplied convergence function

do_bottom(ts, chi)[source]¶

Perform bottom CTM move

Parameters

ts (adpeps.utils.ctmtensors.CTMTensors) – input tensors
chi (int) – boundary bond dimension

Returns

updated tensors

Return type

CTMTensors

do_left(ts, chi)[source]¶

Perform left CTM move

Parameters

ts (adpeps.utils.ctmtensors.CTMTensors) – input tensors
chi (int) – boundary bond dimension

Returns

A tuple containing

tensors (CTMTensors): updated tensors
sl (np.ndarray): singular values of C1 (for convergence)

Return type

Tuple[CTMTensors, ndarray]

do_right(ts, chi)[source]¶

Perform right CTM move

Parameters

ts (adpeps.utils.ctmtensors.CTMTensors) – input tensors
chi (int) – boundary bond dimension

Returns

updated tensors

Return type

CTMTensors

do_top(ts, chi)[source]¶

Perform top CTM move

Parameters

ts (adpeps.utils.ctmtensors.CTMTensors) – input tensors
chi (int) – boundary bond dimension

Returns

updated tensors

Return type

CTMTensors

get_projectors(T1, T2, chi)[source]¶

Contracts the corners together and computes the projectors by performing an svd

Parameters: T1 (int) –

get_projectors_bottom(ts, chi)[source]¶

Returns the bottom projectors

Return type

Tuple[ndarray, ndarray, ndarray]

Parameters

ts (adpeps.utils.ctmtensors.CTMTensors) –
chi (int) –

get_projectors_left(ts, chi)[source]¶

Returns the left projectors

Return type

Tuple[ndarray, ndarray, ndarray]

Parameters

ts (adpeps.utils.ctmtensors.CTMTensors) –
chi (int) –

get_projectors_right(ts, chi)[source]¶

Returns the right projectors

Return type

Tuple[ndarray, ndarray, ndarray]

Parameters

ts (adpeps.utils.ctmtensors.CTMTensors) –
chi (int) –

get_projectors_top(ts, chi)[source]¶

Returns the top projectors

Return type

Tuple[ndarray, ndarray, ndarray]

Parameters

ts (adpeps.utils.ctmtensors.CTMTensors) –
chi (int) –

renorm_left(ts, Pl, Plb)[source]¶

Return type

Tuple[Nested, Nested, Nested]

Parameters

ts (adpeps.utils.ctmtensors.CTMTensors) –
Pl (jax._src.numpy.lax_numpy.ndarray) –
Plb (jax._src.numpy.lax_numpy.ndarray) –

renormalize(tensors, chi)[source]¶

Performs a CTM iteration Updates all sites in the unit cell

Parameters

tensors (adpeps.utils.ctmtensors.CTMTensors) – input ctmtensors
chi (int) – boundary bond dimension

Returns

A tuple containing

tensors (CTMTensors): updated tensors
S (np.ndarray): singular values of C1 (for convergence)

Return type

Tuple[CTMTensors, ndarray]