Releases: google/TensorNetwork
Releases · google/TensorNetwork
bug fixes
MPS improvements, bug fixes
Bug fixes in BaseMPS class.
Bug fixes and minor improvements
0.4.4 update description
NconBuilder
Added the NconBuilder class.
General improvements.
- Added
tn.from_topology
method - Improved JAX backend, especially around decomposition algorithms
- Added many features to
tn.Tensor
- Added new encoding for block sparse backend
Bug fixes and optimizations
A minor release that doesn't change user facing behavior. Many methods were optimized and several minor bugs were fixed.
DMRG Support
DRMG is now supported on the NumPy, JAX, and PyTorch backends.
- Added
tn.FiniteDMRG
class - Added
tn.FiniteMPO
along with several prebuilt MPO classes likeFiniteTFI
andFiniteXXZ
. - JAX backend now
jit
s several of its methods. - Added experimental
tn_keras
layers. These layers are HIGHLY EXPERIMENTAL, so please don't run them in production yet. :) - Added a new
tn.get_neighbors
method. - Added new
ZN
symmetry group for thesymmetric
backend.
0.3.1
Minor release for backend bug fix.
Support for symmetries
Added a new symmetric
backend. Current symmetries include U1 and Z2, but more will be added in the very near future.
- Added various classes for the symmetric backend including:
BlockSparseTensor
,ChargeArray
,Index
,U1Charge
,Z2Charge
, andBaseCharge
. - Added
tn.replicate_nodes
as a cleaner API totn.copy()
Inifinte MPS
- Added
InifiniteMPS
class. - Added
+
,-
,*
, and/
support forNode
s - Added
with tn.DefaultBackend(....):
support to allow more modular control of the default backends. - Added various quantum abstractions like
QuOperator
,QuVector
andQuAdjointVector
. - Added support for contracting disconnected graphs in all the
contractors
. - Added support for creating/passing a custom backend.