The Apache Mahout team is pleased to announce the release of qumat 0.6.0. qumat is a library for composing quantum machine learning circuits across multiple backends (Qiskit, Cirq, Amazon Braket). The optional qumat[qdp] extra provides GPU-accelerated quantum data processing.
Install: pip install qumat==0.6.0 pip install "qumat[qdp]==0.6.0" # Linux x86_64 + NVIDIA CUDA Project page: https://mahout.apache.org/ PyPI: https://pypi.org/project/qumat/0.6.0/ Source/Git: https://github.com/apache/mahout/releases/tag/mahout-qumat-0.6.0 This release rolls up 111 PRs since 0.5.0. Highlights: - **QDP encoding parity and new encodings.** Phase, IQP, and IQP-Z encodings now ship on both NVIDIA CUDA and AMD ROCm backends, closing parity gaps with the existing angle and amplitude paths. New float32 zero-copy batch paths and GPU-pointer encoding for the IQP family cut host-device copies in hot loops. - **AMD GPU support.** New Mahout-AMD framework with PennyLane-AMDGPU integration; AMD ROCm is now a first-class backend selectable from the QDP encoding and throughput benchmarks. CUDA kernel build targets are configurable for forward compatibility with newer architectures. - **Benchmarks and documentation overhaul.** New SVHN Quantum Kernel SVM and IQP latency/throughput benchmarks. Docs site gained frontmatter across all pages, a troubleshooting guide, self-hosted KaTeX (offline-friendly), and CONTRIBUTING merged into the README for easier onboarding. We welcome your feedback on the [email protected] mailing list. Regards, Ryan Huang Release Manager, Apache Mahout
