Tensorcircuit

Added

- add gradient free scipy interface for optimization

- add qiskit circuit to tensorcircuit circuit methods

- add draw method on circuit from qiskit transform pipeline

Changed

- futher refactor VQNHE code in applications

- add alias `sample` for `perfect_sampling` method

- optimize VQNHE pipeline for a more stable training loop (breaking changes in some APIs)

Fixed

- Circuit inputs will convert to tensor first

Added

- add sigmoid method on backends

- add MPO expectation template function for MPO evaluation on circuit

- add `operator_expectation` in templates.measurements for a unified expectation interface

- add `templates.chems` module for interface between tc and openfermion on quantum chemistry related tasks

- add tc.Circuit to Qiskit QuantumCircuit transformation

Fixed

- fix the bug in QuOperator.from_local_tensor where the dtype should always be in numpy context

- fix MPO copy when apply MPO gate on the circuit

Changed

- allow multi-qubit gate in multicontrol gate

Added

- new universal contraction analyse tools and pseudo contraction rehearsals for debug

- add `gather1d` method on backends for 1d tensor indexing

- add `dataset` module in template submodule for dataset preprocessing and embedding

- MPO format quantum gate is natively support now

- add multicontrol gates in MPO format

Fixed

- fixed real operation on some methods in templates.measurements

Changed

- add gatef key in circuit IR dict for the gate function, while replace gate key with the gate node or MPO (breaking change)

Added

- add `td` and `sd` gates for dagger version of T gate and S gate

- add `argmax` and `argmin` as backend methods

- add `expectation_before` methods for `tc.Circuit` for further manipulation on the tensornetwork

Changed

- refined repr for `tc.gates.Gate`

- expectation API now supports int index besides list indexes

Fixed

- make consistent `Gate` return for channels

- fixed bug on list optimizer for contraction

- stability for QR operator in terms of automatic differentiation

Added

- add `hessian` method on backends

- add further automatic pipelines for visualization by generating pdf or images

- add `reshape2` method on backends as a short cut to reshape a tensor with all legs 2-d

- add `reshapem` method on backends to reshape any tensor as a square matrix

- add `controlled` and `ocontrolled` API to generate more gates

- add `crx`, `cry`, `crz` gate on `Circuit`

- add `__repr__` and `__str__` for backend object

- `tc.expectation` now support ket arg as quvector form

Fixed

- `sizen` correctly returns 1 for tensor of no shape

- fixed `convert_to_tensor` bug in numpy backend in TensorNetwork

- `any_gate` also support Gate format instead of matrix

- `matrix_for_gate` works now for backends more than numpy

Changed

- `expectation` API now also accepts plain tensor instead of `tc.Gate`.

- `DMCircuit` and `DMCircuit2` are all pointing the efficent implementations (breaking changes)

Added

- add `solve` method on backends to solve linear equations

- add full quantum natural gradient examples and `qng` method in experimental module

- add `concat` method to backends

- add `stop_gradient` method to backends

- add `has_aux` arg on `vvag` method

- add `imag` method on backends

- add `Circuit.vis_tex` interface that returns the quantikz circuit latex

Changed

- contractor, dtype and backend set are default to return objects, `with tc.runtime_backend("jax") as K` or `K = tc.set_backend("jax")` could work

- change `perfect_sampling` to use `measure_jit` behind the scene

- `anygate` automatically reshape the unitary input to 2-d leg for users' good

- `quantum.renyi_entropy` computation with correct prefactor

- `Circuit` gate can provided other names by name attr

- `example_block` support param auto reshape for users' good

Fixed

- make four algorithms for quantum natural gradient consistent and correct

- torch `real` is now a real