Tensorcircuit

Added

- Add `device` and `device_move` methods on backend with universal device string representation

- Add `to_qir`, `from_qir`, `to_circuit`, `inverse`, `from_qiskit`, `to_qiskit`, `sample`, `prepend`, `append`, `cond_measurment`, `select_gate` method for `DMCircuit` class

- Add `status` arguments as external randomness for `perfect_sampling` and `measure` methods

- `DMCircuit` now supports `mps_inputs` and `mpo_dminputs`

- Add decorator `tc.interfaces.args_to_tensor` to auto convert function inputs as tensor format

Changed

- Refactor circuit and dmcircuit with common methods now in `basecircuit.py`, and merge dmcircuit.py and dmcircuit2.py, now `DMCircuit` supports MPO gate and qir representation

Fixed

- Patially solve the issue with visualization on `cond_measure` (50)

Added

- PyTorch backend support multi pytrees version of `tree_map`

- Add `dtype` backend method which returns the dtype string

- Add TensorFlow interface

- Add `to_dlpack` and `from_dlpack` method on backends

- Add `enable_dlpack` option on interfaces and torchnn

- Add `inverse` method for Circuit (26)

Changed

- Refactor `interfaces` code as a submodule and add pytree support for args

- Change the way to register global setup internally, so that we can skip the list of all submodules

- Refactor the tensortrans code to a pytree perspective

Fixed

- Fixed `numpy` method bug in pytorch backend when the input tensor requires grad (24) and when the tensor is on GPU (25)

- Fixed `TorchLayer` parameter list auto registeration

- Pytorch interface is now device aware (25)

Added

- Add `enable_lightcone` option in circuit `expectation` method, where only gates within casual lightcone of local observable is contracted.

- Add `benchmark` function into utils

Fixed

- Fixed a vital bug on circuit expectation evaluation, a wrongly transposed operator connection is fixed.

- Name passed in gate application now works as Node name

Added

- Add PyTorch nn Module wrapper in `torchnn`

- Add `reverse`, `mod`, `left_shift`, `right_shift`, `arange` methods on backend

- Brand new `sample` API with batch support and sampling from state support

- add more methods in global namespace, and add alias `KerasLayer`/`TorchLayer`

Fixed

- Fixed bug in merge single gates when all gates are single-qubit ones

Changed

- The default contractor enable preprocessing feature where single-qubit gates are merged firstly

Added

- Add more type auto conversion for `tc.gates.Gate` as inputs

- Add `tree_flatten` and `tree_unflatten` method on backends

- Add torch optimizer to the backend agnostic optimizer abstraction

Changed

- Refactor the tree utils, add native torch support for pytree utils

Fixed

- grad in torch backend now support pytrees

- fix float parameter issue in translation to qiskit circuit (19)

Added

- Add `rxx`, `ryy` and `rzz` gate

Fixed

- Fix installation issue with tensorflow requirements on MACOS with M1 chip

- Improve M1 macOS compatibility with unjit tensorflow ops

- Fixed SVD backprop bug on jax backend of wide matrix

- `mps_input` dtype auto correction enabled