tf-quant-finance

TF Quant Finance is an archived, unmaintained Python library for quantitative finance built on TensorFlow. It is designed to use TensorFlow's hardware acceleration, automatic differentiation, vectorization, and XLA compilation. The library is organized in three tiers: 1. Foundational mathematical methods: optimization, interpolation, root finders, linear algebra, and random or quasi-random number generation. 2. Mid-level methods: ODE and PDE solvers, an Ito process framework, diffusion path generators, and copula samplers. 3. Quant-finance utilities: specific pricing models such as Local Volatility, Stochastic Volatility, Stochastic Local Volatility, and Hull-White; model calibration; rate-curve building; payoff descriptions; and schedule generation. The documented roadmap also listed Brownian motion, geometric Brownian motion, Ornstein-Uhlenbeck, one-factor Hull-White, Heston, local volatility, quadratic local volatility, and SABR models, along with Dupire and SABR calibration, Hagan-West yield-curve bootstrapping, Monotone Convex interpolation, and support for dates, day-count conventions, and holidays. Installation is through pip, requiring Python 3.7 and TensorFlow 2.7 or newer. The repository includes Jupyter notebook examples for American option pricing under Black-Scholes, Euler-scheme Monte Carlo, Black-Scholes prices and implied volatility, forward and backward gradients, Brent root search, optimization, swap-curve fitting, and vectorization/XLA. Development uses Bazel for tests and building a custom pip wheel. The project is licensed under Apache 2.0, while the included Sobol primitive polynomials and initial direction numbers use a BSD license. Because the repository is archived, maintainers recommend forking it if continued development is needed.