You signed in with another tab or window. Reload to refresh your session.You signed out in another tab or window. Reload to refresh your session.You switched accounts on another tab or window. Reload to refresh your session.Dismiss alert
Before #282, Lattice (and thus Ansatz) would use Sites to locate tensors. This is problematic when working with Operators (like MPO) or dual states (like the adjoint of a MPS) because either they don't posses those Sites or they
#282 introduces the concept of Lane, a Site without i/o information (well, it would be more correct to say that "a Site is a Lane with i/o information"), and forces Ansatz and its subtypes to use Lane to localize Tensors
Changes to AbstractQuantum
Stop supporting tensors(; at) at this level
inds(; at::Site) is still supported
in the rare case is needed, use tensors(; contains=inds(; at))
Changes to AbstractAnsatz
Add a mapping between Lane and Tensor
Add a mapping between Bond and Symbol of the indices
Check consistency between the Lattice and the TensorNetwork
Should run on many operations automatically
Add sth similar to @unsafe_region but for AbstractAnsatz
It should stop running the check, run it in the end of the scope and activate them again
Furthermore, it might be interesting to use some kind of reactive programming to automatically propagate changes performed to the innermost layers (i.e. TensorNetwork and Quantum for Ansatz) to the parent layers. If not, we must intercept many calls to the more generic versions.
The text was updated successfully, but these errors were encountered:
Before #282,
Lattice
(and thusAnsatz
) would useSite
s to locate tensors. This is problematic when working withOperator
s (likeMPO
) or dual states (like the adjoint of aMPS
) because either they don't posses thoseSite
s or they#282 introduces the concept of
Lane
, aSite
without i/o information (well, it would be more correct to say that "aSite
is aLane
with i/o information"), and forcesAnsatz
and its subtypes to useLane
to localizeTensor
sChanges to
AbstractQuantum
tensors(; at)
at this levelinds(; at::Site)
is still supportedtensors(; contains=inds(; at))
Changes to
AbstractAnsatz
Lane
andTensor
Bond
andSymbol
of the indicesLattice
and theTensorNetwork
@unsafe_region
but forAbstractAnsatz
Furthermore, it might be interesting to use some kind of reactive programming to automatically propagate changes performed to the innermost layers (i.e.
TensorNetwork
andQuantum
forAnsatz
) to the parent layers. If not, we must intercept many calls to the more generic versions.The text was updated successfully, but these errors were encountered: