Tensorcircuit

Added

- add `quantum.heisenberg_hamiltonian` for hamiltonian generation shortcut

- add `has_aux` parameter in backend methods `grad` and `value_and_grad`, the semantic syntax is the same as jax

- add `optimizer` class on tensorflow and jax backend, so that a minimal and unified backend agnostic optimizer interface is provided

- add `quantum.mutual_information`, add support on mixed state for `quantum.reduced_density_matrix`

- add `jvp` methods for tensorflow, jax, torch backends, and ensure pytree support in `jvp` and `vjp` interfaces for tensorflow and jax backends; also ensure complex support for `jvp` and `vjp`

- add `jacfwd` and `jacrev` for backend methods (experimental API, may have bugs and subject to changes)

Fixed

- fix `matmul` bug on tensornetwork tensorflow backend

Changed

- delete `qcode` IR for `Circuit`, use `qir` instead (breaking changes)

- basic circuit running is ok on pytorch backend with some complex support fixing

Added

- add `get_random_state` and `random_split` methods to backends

- add qir representation of circuit, `c.to_qir()` and `Circuit.from_qir()` methods

- fine-grained control on backend, dtype and contractor setup: `tc.set_function_backend()` for function level decorator and `tc.runtime_backend()` as with context manager

- add `state_centric` decorator in `tc.templates.blocks` to transform circuit-to-circuit funtion to state-to-state function

- add `interfaces.scipy_optimize_interface` to transform quantum function into `scipy.optimize.minimize` campatible form

Fixed

- avoid error on watch non `tf.Tensor` in tensorflow backend grad method

- circuit preprocessing simplification with only single qubit gates

- avoid the bug when random from jax backend with jitted function

- refresh the state cache in Circuit when new gate is applied

Changed

- refactor `tc.gates` (breaking API on `rgate` -> `r_gate`, `iswapgate` -> `iswap_gate`)


TensorCircuit is initially a personal project created by refraction-ray (Shi-Xin Zhang). He began this project in April 2020, inspired by the MPS quantum simulator [mpsim](https://github.com/grmlarose/mpsim) and the introduction of the Google [TensorNetwork](https://github.com/google/TensorNetwork) package. This project is further developed by him during 2020 and the first half of 2021 when he was a Ph.D. candidate at Tsinghua University, with multiple new features and applications added for his research purpose. The original TensorCircuit project is archived now on [GitHub](https://github.com/refraction-ray/tensorcircuit/). He decided to make this project a more universal open-source framework after he joined Tencent in July 2021. And he has extensively refactored and optimized the codebase since then. As the creator and the lead author of TensorCircuit, he thanks all the [contributors](https://github.com/tencent-quantum-lab/tensorcircuit#contributors) who have made TensorCircuit and the ecosystem better.