Skip to content

Release 0.39.0

Choose a tag to compare

View all tags
@multiphaseCFD multiphaseCFD released this 05 Nov 14:37
· 250 commits to master since this release
v0.39.0
dcea959

New features since last release

  • Add support for out-of-order qml.probs in lightning.gpu. (#941)

  • Add mid-circuit measurements support to lightning.gpu's single-GPU backend. (#931)

  • Integrate Lightning-GPU with Catalyst so that hybrid programs can be seamlessly QJIT-compiled and executed on this device following pip install pennylane-lightning-gpu. (#928)

  • Add qml.Projector observable support via diagonalization to Lightning-GPU. (#894)

  • Add 1-target wire controlled gate support to lightning.tensor. Note that cutensornet only supports 1-target wire controlled gate as of v24.08. A controlled gate with more than 1 target wire should be converted to dense matrix. (#880)

  • Build and upload Lightning-Tensor wheels (x86_64, AARCH64) to PyPI. (#862) (#905)

  • Add Matrix Product Operator (MPO) for all gates support to lightning.tensor. Note current C++ implementation only works for MPO sites data provided by users. (#859)

  • Add shots measurement support to lightning.tensor. (#852)

  • Lightning-GPU and Lightning-Kokkos migrated to the new device API. (#853) (#810)

Breaking changes

  • Update MacOS wheel build to 13.0 for X86_64 and ARM due to the deprecation of MacOS-12 CI runners. (#969)

  • Deprecate initSV() and add resetStateVector() from the C++ API Lightning-GPU. This is to remove the reset_state additional call in the Python layer. (#933)

  • Deprecate PI gates implementation in Lightning-Qubit. The PI gates were the first implementation of gate kernels in lightning.qubit using pre-computed indices, prior to the development of LM (less memory) and AVX kernels. This deprecation is in favour of reducing compilation time and ensuring that Lightning-Qubit only relies on LM kernels in the dynamic dispatcher across all platforms. (#925)

  • Remove PowerPC wheel build recipe for Lightning-Qubit. (#902)

  • Update MacOS wheel builds to require Monterey (12.0) or greater for x86_64 and ARM. This was required to update Pybind11 to the latest release (2.13.5) for enabling Numpy 2.0 support in Lightning. (#901)

  • Remove support for Python 3.9 for all Lightning simulators. (#891)

Improvements

  • Update the lightning.tensor Python layer unit tests, as lightning.tensor cannot be cleaned up like other state-vector devices because the data is attached to the graph. It is recommended to use one device per circuit for lightning.tensor. (#971)

  • Add joint check for the N-controlled wires support in lightning.qubit. (#949)

  • Optimize GlobalPhase and C(GlobalPhase) gate implementation in lightning.gpu. (#946)

  • Add missing liblightning_kokkos_catalyst.so when building Lightning-Kokkos in editable installation. (#945)

  • Optimize the cartesian product to reduce the amount of memory necessary to set the StatePrep in Lightning-Tensor. (#943)

  • Update the qml.probs data-return in Lightning-GPU C++ API to align with other state-vector devices. (#941)

  • Add zero-state initialization to both StateVectorCudaManaged and StateVectorCudaMPI constructors to remove the reset_state in the Python layer ctor and refactor setBasisState(state, wires) in the C++ API. (#933)

  • Add setStateVector(state, wire) support to the Lightning-GPU C++ API. (#930)

  • The generate_samples methods of lightning.qubit and lightning.kokkos can now take in a seed number to make the generated samples deterministic. This can be useful when, among other things, fixing flaky tests in CI. (#927)

  • Remove dynamic decomposition rules for all Lightning devices. (#926)

  • Always decompose qml.QFT in all Lightning devices. (#924)

  • Uniform Python format to adhere PennyLane style. (#924)

  • Add the ci:use-gpu-runner GitHub label to lightning.kokkos GPU Testing CIs. (#916)

  • Update the test suite to remove deprecated code. (#912)

  • Merge lightning.gpu and lightning.tensor GPU tests in single Python and C++ CIs controlled by the ci:use-gpu-runner label. (#911)

  • Skip the compilation of Lightning simulators and development requirements to boost the build of public docs up to 5x. (#904)

  • Build Lightning wheels in Release mode to reduce the binary sizes. (#903)

  • Update Pybind11 to 2.13.5. (#901)

  • Migrate wheels artifacts to v4. (#893)

  • Update GitHub actions in response to a high-severity vulnerability. (#887)

  • Optimize and simplify controlled kernels in Lightning-Qubit. (#882)

  • Optimize gate cache recording for Lightning-Tensor C++ API. (#879)

  • Unify Lightning-Kokkos and Lightning-Qubit devices under a Lightning-Base abstracted class. (#876)

  • Smarter defaults for the split_obs argument in the serializer. The serializer splits linear combinations into chunks instead of all their terms. (#873)

  • Prefer tomlkit over toml for building Lightning wheels, and choose tomli and tomllib over toml when installing the package. (#857)

  • Lightning-Kokkos gains native support for the PauliRot gate. (#855)

Documentation

  • Update Lightning-Tensor installation docs and usage suggestions. (#971) (#972)

  • Update README.rst installation instructions for lightning.gpu and lightning.tensor. (#957)

  • Update lightning.tensor documentation to include all the new features added since pull request #756. The new features are: 1. Finite-shot measurements; 2. Expval-base quantities; 3. Support for qml.state() and qml.stateprep(); 4. Support for all gates support via Matrix Product Operator (MPO). (#909)

Bug fixes

  • Fix Lightning Kokkos test_device for kokkos_args fail for MacOS due to np.complex256 (#974)

  • Fix PTM stable-latest related to default.qubit.legacy and the latest flag usage. (#961) (#966)

  • Fix build failure for Lightning-Kokkos editable installation on MacOS due to liblightning_kokkos_catalyst.so copy and liblightning_kokkos_catalyst.so not copied to correct build path for editable installation. (#947) (#968)

  • Add concept restriction to ensure ConstMult inline function only hit with arithmetic-values times complex values. Fixes build failures with the test suite when enabling OpenMP, and disabling BLAS and Python under clang. (#936)

  • Bug fix for applyMatrix in Lightning-Tensor. Matrix operator data is not stored in the cuGateCache object to support TensorProd obs with multiple Hermitian obs. (#932)

  • Bug fix for _pauli_word of QuantumScriptSerializer. _pauli_word can process PauliWord object: I. (#919)

  • Bug fix for analytic qml.probs in the Lightning-Tensor C++ API. (#906)

Contributors

This release contains contributions from (in alphabetical order):

Ali Asadi, Amintor Dusko, Diego Guala, Joseph Lee, Luis Alfredo Nuñez Meneses, Vincent Michaud-Rioux, Lee J. O'Riordan, Mudit Pandey, Shuli Shu, Haochen Paul Wang

Assets 46
Loading