Release 0.42.0

New features since last release

• Added distributed simulation using MPI to the lightning.kokkos device,
  enabling larger quantum circuits by distributing workloads across multiple CPUs or GPUs. (#1114)

• Mid-circuit measurements using the tree-traversal algorithm are now supported
  in the lightning.qubit, lightning.kokkos and lightning.gpu devices,
  providing both significant memory savings and sampling efficiency! (#1166)

Improvements 🛠

• Lightning devices can now accept a seed argument to enable deterministic shots measurements. (#1171)

• PennyLane-Lightning is compatible with JAX version 0.5.3+. (#1152) (#1161)

• Refactored MPIManager class from lightning.gpu to base class. (#1162)

Breaking changes 💔

• The LightningBaseStateVector, LightningBaseAdjointJacobian, LightningBaseMeasurements,
  LightningInterpreter and QuantumScriptSerializer base classes now can be found at pennylane_lightning.lightning_base.
  The new lightning_base module further enables the relocation of core files
  from pennylane_lightning/core/src/* to pennylane_lightning/core/*.
  The license classifier and project.license as a TOML table are deprecated in favor of a SPDX license
  expression and removed in pyproject.toml.
  To speedup the recompilation of C++ source code, ccache is also added to Makefile. (#1098)

• Updated pyproject to adhere to PyPA binary distribution format for built wheels. (#1193)

Deprecations 👋

• Added and subsequently removed native support for MultiControlledX gates in lightning.tensor. (#1169) (#1183)

Documentation 📝

• Converted the README to markdown (as in the pennylane and catalyst repositories), and added a header image. (#1139) (#1141) (#1142)

• Edited and prepared CHANGELOG.md for the release v0.42.0. (#1199) (#1204)

• Fixed typo in lightning.gpu documentation adjoint-jacobian section. (#1179)

Bug fixes 🐛

• Fixed the implementation of multi-controlled gates with a single target wire for arbitrary control values in lightning.tensor. (#1169)

• Updated requirements to only download JAX version 0.5.3 for non-X86 MacOS. (#1163)

• Fixed Docker build for lighting.kokkos with ROCM library for AMD GPUs. Updated ROCM from 5.7 to 6.2.4. (#1158)

• Fixed use of Torch with AmplitudeEmbedding by applying qml.broadcast_expand before decomposition in the preprocessing stage. (#1175)

Internal changes ⚙️

• Updated tests with device(..., shots=...) to use qml.set_shots to ensure compatibility with the latest version of PennyLane. (#1173)

• Fixed applyMultiRZ for lightning.kokkos device to use Kokkos function instead of lambda. (#1194)

• Enabled AmplitudeEmbedding Python tests for lightning.kokkos and lightning.gpu devices. (#1192)

• Updated docker build CI for stable version to use v0.41.1. (#1188)

• Removed flaky tests and added random seeding to measurement tests. (#1172) (#1196)

• Improved CI to use local catalyst repository instead of fetching on Github CI. (#1164)

• Updated the Lightning build dependencies. (#1168)

• Updated dev requirements to use JAX version 0.6.0 for CI tests. (#1161)

• Updated stable requirements to use JAX version 0.4.28 for CI tests. (#1160)

• Bumped readthedocs Github action runner to use Ubuntu-24.04. (#1151)

• Updated the clang-format and clang-tidy versions to v20 for compatibility with Catalyst. (#1153)

• Removed max_workers argument for default.qubit device in Python tests to reduce CI testing time. (##1174)

• Bumped Github action runner to use Ubuntu-24.04 or Ubuntu-latest.
  Fixed all ubuntu-latest action runners to ubuntu-24.04. (#1167)

• Merged the v0.41.0-rc branch to the master and bumped the version. (#1132)

Contributors ✍️

This release contains contributions from (in alphabetical order):

Runor Agbaire,
Ali Asadi,
Amintor Dusko,
Yushao Chen,
Christina Lee,
Joseph Lee,
Mehrdad Malekmohammadi,
Anton Naim Ibrahim,
Luis Alfredo Nuñez Meneses,
Mudit Pandey,
Shuli Shu,
Marc Vandelle,
Jake Zaia

---