Packages
Extends TensorFlow with support for file systems and data formats not built into TensorFlow, including HTTP/HTTPS access and automatic decompression for remote datasets.
Install only if you have a specific need for extended I/O capabilities; it is not required for standard TensorFlow workflows.
Interpolates networks of curves into smooth Gordon surfaces using B-spline mathematics, integrated with CadQuery's OCP for CAD geometry operations.
pytransform3d provides operations, conversions, and visualizations for 3D rotations and translations, with built-in support for multiple rotation representations and a TransformManager for chaining complex transformations.
PyTorch Forecasting provides neural network models and utilities for time series forecasting, including architectures like Temporal Fusion Transformers, N-BEATS, and N-HiTS, with training orchestrated through PyTorch Lightning.
Install it if you need to train neural forecasting models on time series data and want to avoid reimplementing standard architectures or training boilerplate.
Provides Python bindings to NVIDIA's fatbin compiler library for working with CUDA binary formats and GPU code compilation artifacts.
However, verify the license terms before use in proprietary projects, and confirm that your system meets any undocumented CUDA runtime dependencies.
Manages cryptographic keys and wallets for the Bittensor network, providing Python bindings to a Rust-based wallet implementation for creating and accessing blockchain credentials.
Install only if you are maintaining legacy code that explicitly depends on bittensor-wallet.
glum is a high-performance Python library for fitting generalized linear models (GLMs) with support for multiple distributions, regularization methods, formula-based specification, and statistical inference.
Provides efficient matrix classes for tabular data that mix dense, sparse, and categorical columns, optimized for statistical algorithms.
Computes empirical calibration weights to correct bias in survey samples and observational data through convex optimization, supporting weight clipping and inexact calibration.
PyMC3 is a Python package for Bayesian statistical modeling and probabilistic machine learning, implementing advanced Markov chain Monte Carlo (MCMC) and variational inference algorithms for complex probabilistic models.
However, it is now legacy—the PyMC project has been renamed and moved to newer backends.
Computes edit distance, alignment, and word error rate (WER) between sequences using Kaldi's original algorithms, with support for compound word matching and statistical confidence intervals.
QuantStats computes portfolio performance metrics (Sharpe ratio, volatility, drawdown, win rate) and generates visualizations and HTML tearsheets for analyzing investment returns.
pyAgrum is a Python library for creating, learning, and performing inference on Bayesian Networks and other Probabilistic Graphical Models, with a C++ backend and high-level Python interface.
RoPWR computes optimal piecewise polynomial regression with support for monotonic constraints, multiple objective functions, and regularization, using convex optimization solvers.
However, note the aging maintenance status (last commit January 2026, 203 days ago) and modest popularity (top 15000 tier)—suitable for niche regression tasks but not…
OptBinning discretizes numeric variables into bins optimally for binary, continuous, and multiclass targets using mathematical programming, with support for constraints.
However, verify that its mathematical programming approach and dependency footprint fit your performance and deployment constraints before committing to production use.
cvc5 is a satisfiability modulo theories (SMT) solver that determines whether first-order logic formulas are satisfiable under combinations of theories; it provides Python bindings to use the solver as a library.
Transforms bilinear and linear forms into sparse matrices and vectors for finite element assembly, supporting 1D, triangular, quadrilateral, tetrahedral, and hexahedral elements plus specialized element types.
Manipulates complex, non-rectangular data structures (jagged arrays, nested records, heterogeneous types, sparse or masked data) using Numpy-like vectorized operations instead of Python loops.
PySAT wraps modern Boolean satisfiability solvers and provides cardinality and pseudo-Boolean encodings, letting you call SAT oracles from Python code.
Not recommended if you only need a pure-Python SAT solver or have no need for low-level solver integration.
Provides timelock encryption for Bittensor commit-reveal weight commitments and general-purpose data using drand randomness, with Rust core and Python bindings.
However, the repository is archived and marked abandoned, meaning no active maintenance or bug fixes should be expected.
Provides synchronous and asynchronous rate limiters to regulate loop execution frequency at a specified rate, similar to rospy.Rate.
Install it if you need straightforward loop frequency regulation in synchronous or asynchronous code.
Adds physics-oriented methods to ROOT file classes, primarily TLorentzVectors and histograms, as a companion library to uproot 3.x for working with particle physics data.
Reads and writes ROOT files (the data format used in high-energy physics) using pure Python and NumPy, without requiring the C++ ROOT library.
Computes the cumulative distribution function of a weighted sum of chi-squared random variables using four moment-matching approximation methods: Hall-Buckley-Eagleson, Satterthwaite-Welch, Wood's F, and Lindsay-Pilla-Basak.
TensorFlow Probability provides probabilistic modeling, statistical inference, and Bayesian machine learning tools integrated with TensorFlow and JAX, including distributions, bijectors, inference algorithms, and probabilistic neural network layers.
dimod provides classes and interfaces for defining and solving binary quadratic models (BQMs), including Ising and QUBO formulations, along with reference samplers and abstract base classes for building custom solvers.
Install it if you need a standard interface for BQM solvers or plan to use quantum annealing tools.
Galois extends NumPy arrays to perform arithmetic operations over finite fields (Galois fields), enabling efficient computation in GF(p^m) with a NumPy-compatible interface.
The non-constant-time implementation is explicitly documented as unsuitable for production security but appropriate for research and education, so install only if…
Fits continuous piecewise linear functions to data by specifying the number of line segments and providing x-y data, with support for weighted fits and multiple optimization strategies.
Install it if you need to model data with distinct linear regimes or approximate nonlinear relationships with continuous piecewise linear segments.
Distributes weighted items into bins using greedy algorithms, either to a fixed number of bins with balanced loads or to the minimum bins with a maximum capacity constraint.
Install it if you need to distribute weighted items across fixed bins or minimize bin count.
Haiku is a neural network library for JAX that provides an object-oriented module abstraction and function transformation to build and train models while retaining access to JAX's pure function transformations.
Performs fast sparse matrix multiplication with integrated top-n result selection, reducing memory footprint and improving performance over separate multiplication and filtering steps.
Extends PyMC with specialized distributions, alternative inference methods, and statespace models that are too specialized for the core library but useful enough to avoid reimplementing.
Install it if you need specialized distributions, alternative inference methods, or statespace models beyond what PyMC core offers.
tf-nightly-cpu provides a CPU-optimized nightly build of TensorFlow for numerical computation, machine learning, and deep learning across CPUs, GPUs, TPUs, and edge devices.
PyTensor-distributions provides a collection of probability distribution functions (pdf, cdf, ppf, logpdf, etc.) and statistical properties (mean, variance, entropy, etc.) implemented as PyTensor expressions for use in probabilistic computing.
However, the alpha status means the API and feature coverage may change; verify that the distributions and expressions you need are already implemented before…
Converts machine learning models to LiteRT format for deployment on mobile and embedded devices with low latency and minimal binary size.
Pynini compiles grammar rules into weighted finite-state transducers and related automata structures, enabling programmatic construction and application of complex linguistic transformations.
RoMa provides differentiable conversions between 3D rotation representations (rotation vectors, quaternions, rotation matrices, Euler angles) and utilities for rotation-space operations in PyTorch, designed for machine learning and gradient-based optimization.
Install it if your project involves 3D rotations, pose estimation, or rotation-aware neural networks.
BlackJAX provides a collection of MCMC and other sampling algorithms built on JAX that run on CPU and GPU, designed for users who have a log-probability function and need a performant sampler.
iminuit is a Python interface to MINUIT2 (CERN's C++ optimization library) for fitting statistical models via maximum-likelihood and least-squares methods, with built-in cost functions and Jupyter visualization.
cvxpy-base provides compiled solver kernels for CVXPY, a Python modeling language for convex optimization problems. It handles the numerical computation layer that solves optimization problems expressed in CVXPY's high-level syntax.