Packages
Provides NVIDIA's CUDA C++ Core Compute Libraries (CCCL) for GPU-accelerated computing on CUDA 12 platforms.
Python MIP models and solves mixed-integer linear programs (MIPs) using a high-level syntax with support for advanced solver features like cut generation, lazy constraints, and solution pools.
Python interface to the igraph graph library for network analysis and complex graph algorithms; this is a legacy package superseded by the igraph package.
Solves strictly convex quadratic programming problems using the Goldfarb/Idnani dual algorithm, minimizing a quadratic objective subject to linear inequality constraints.
POT provides solvers for optimal transport problems, including Wasserstein distances, Gromov-Wasserstein distances, and related algorithms for signal processing, image processing, and machine learning applications.
Gin provides a lightweight configuration framework that lets you decorate Python functions and classes to accept parameter values from config files or command-line arguments, eliminating the need for manual configuration object management.
Install it if parameter management is a pain point in your workflow.
Snuggs evaluates s-expressions (Lisp-like syntax) that operate on NumPy arrays and scalars, letting you write array computations as nested function calls rather than Python code.
Not recommended if you need optimized array computation; use numexpr or raw NumPy instead.
Trains and deploys decision forest models (Random Forests, Gradient Boosted Trees) within TensorFlow for classification, regression, and ranking tasks.
Not recommended for Windows users without WSL, or for projects not already committed to TensorFlow.
geoh5py provides a Python API to read, write, and manipulate geoscientific data (points, curves, surfaces, grids) in the geoh5 file format, enabling interoperability with Geoscience ANALYST visualization software.
Install it if you work with geoh5 data or need to integrate geoscientific datasets with Geoscience ANALYST.
Parses LaTeX math expressions and converts them to SymPy symbolic form, supporting arithmetic, functions, calculus, linear algebra, and matrix operations.
However, be aware that the last release was in 2023 and there have been no recent commits, so bug fixes or updates for newer Python versions are unlikely.
Device-side API extensions for performing mathematical calculations inside CUDA kernels, enabling fused numerical operations to reduce latency and improve application performance.
Pingouin provides a comprehensive statistical analysis toolkit for Python, offering t-tests, ANOVAs, correlations, regression, effect sizes, power analysis, and hypothesis testing with detailed output including confidence intervals and Bayes factors.
Install it if you do statistical testing in Python and want richer output than SciPy alone provides.
Lineax solves linear systems and least-squares problems in JAX, handling both explicit matrices and implicit linear operators without materializing them.
Triangle wraps Jonathan Richard Shewchuk's C library to generate two-dimensional quality meshes and compute Delaunay triangulations from point sets or polygonal regions.
However, be aware the project is dormant—maintenance is minimal, so plan accordingly if you need active support or encounter bugs.
Python wrappers for SymEngine, a fast C++ symbolic manipulation library, enabling symbolic computation and algebraic operations in Python.
NumKong provides mixed-precision linear algebra and distance kernels with automatic accumulator widening, GIL-free batched operations, and low-precision dtype support (BFloat16, Float8, Float6, packed bits) across x86, ARM, RISC-V, and other architectures.
MagicCube implements NxNxN Rubik cubes with fast rotations and includes a solver for 3x3x3 cubes and a move optimizer.
Computes numerical derivatives, gradients, Jacobians, Hessians, and higher-order partial derivatives of scalar and vector functions using adaptive finite differences and Richardson extrapolation.
Implements the Munkres algorithm (Hungarian algorithm) to solve the Assignment Problem, finding optimal one-to-one matchings between sets given a cost matrix.
However, the abandoned status (no updates since 2020-09-15) means you should verify the implementation meets your accuracy and performance needs before adopting it in…
XProf is a profiler for ML workloads that visualizes performance across multiple devices, showing execution timelines, memory usage, and operation-level details through a web interface.
DAQP is a dual active-set solver for convex quadratic programs, including mixed-integer variants, with a C-based implementation that interfaces to Python.
Install only if your problem fits the standard convex QP form; for general nonlinear optimization or exotic constraints, a more general solver may be necessary.
Provides C++ implementations of multiple tokenizers (SentencePiece, Tiktoken, Huggingface, Llama2.c, Tekken) with Python bindings for use in PyTorch-based LLM applications and ExecuTorch runners.
ExecuTorch exports and runs PyTorch models on mobile, embedded, and edge devices with support for hardware acceleration backends like XNNPACK, Core ML, and QNN.
YDF trains, evaluates, and interprets decision forest models (Random Forests, Gradient Boosted Trees, and others) through a Python wrapper around the C++ Yggdrasil Decision Forests library.
Solves unconstrained and linearly constrained nonlinear minimization problems without computing derivatives, using direct search methods that work well when objective function evaluations are expensive or noisy.
MOSEK solves large-scale convex and mixed-integer optimization problems including linear, quadratic, conic, and semidefinite programs via a Python API.
Solves for unknown sides and angles of a triangle given partial information, using the law of sines and law of cosines.
However, the package is abandoned and receives no maintenance; do not rely on it for production systems or expect updates.
s2sphere provides Python bindings to work with S2 geometry—a spherical geometry library for spatial indexing and geometric calculations on the Earth's surface.
However, abandonment since 2017-11-16 means no bug fixes, security patches, or compatibility updates are forthcoming—verify it runs on your target Python version…
Adds support for rendering mathematical formulas in Markdown documents by extending Python-Markdown to recognize math delimiters and pass them to MathJax or KaTeX for display.
Install it if you need to render math formulas in Markdown and are willing to load MathJax or KaTeX separately in your HTML output.
Provides pre-built OpenBLAS shared libraries and headers for use as a build or runtime dependency in NumPy, SciPy, and other Python projects that need linear algebra acceleration.
Dask-awkward integrates Awkward Array with Dask to enable distributed, lazy computation on complex, nested data structures across multiple cores or machines.
Install it if you work with Awkward Arrays at scale or need lazy evaluation of nested structures across multiple workers.
PreliZ helps you specify prior distributions for Bayesian models by providing interactive tools for prior elicitation, from parameter-space methods to predictive elicitation, designed to work with probabilistic programming languages like PyMC.
bitmath converts and performs arithmetic on file sizes across SI and NIST prefix units (kB to YiB), with support for human-readable formatting, rich comparisons, and capacity math.
Converts TensorFlow, Keras, TensorFlow.js, and TFLite models to ONNX format via command line or Python API, enabling model portability across different inference runtimes.
Optimistix provides nonlinear solvers for root finding, minimization, fixed-point problems, and least-squares fitting, built on JAX with support for autodiff, GPU/TPU acceleration, and PyTree-based state.
Implements multiway number partitioning and bin packing algorithms with support for exact and approximate methods, multiple input formats, and configurable optimization objectives.
However, the aging maintenance status (last release over a year ago, Pre-Alpha classifier) and lack of recent activity suggest treating it as stable but not actively…
JMP provides mixed precision training support for JAX by managing dtype policies and loss scaling to reduce memory bandwidth and improve computational efficiency during neural network training.
The main gotcha is that JAX must be installed separately first with the correct accelerator support; beyond that, JMP is straightforward to integrate into existing…
Integrates PyTorch profiling data with TensorBoard, providing GPU timeline tracing and performance diagnostics for ML workloads through a TensorBoard plugin interface.
However, the last release was in October 2023; verify compatibility with your current PyTorch version before relying on it for new projects.
Command-line interface for managing Bittensor platform operations including wallet creation, subnet registration, delegation, and governance voting.
Diffrax provides numerical solvers for ordinary, stochastic, and controlled differential equations in JAX, with automatic differentiation and GPU acceleration built in.