# google-research/torchsde

Differentiable SDE solvers with GPU support and efficient sensitivity analysis.

Repository: https://github.com/google-research/torchsde
Canonical: https://ross.abutalabs.com/products/torchsde
Language: Python
License: Apache-2.0
License Family: permissive
Topics: deep-learning, deep-neural-networks, pytorch, dynamical-systems, differential-equations, stochastic-processes, stochastic-differential-equations, stochastic-volatility-models, neural-differential-equations
Archived: true
Last push: 2024-12-30T10:52:51+00:00

## Health v2 (maintenance only)
Score: 10/100 (v2, computed 2026-09-03T02:20:16.233290+00:00)
- activity 0, release rhythm 8, longevity 100
- inputs: {"age_days": 2249, "days_push": 611, "days_rel": null, "gap_med": null, "n_releases_24m": 0}
- flags: archived
- formula: round(0.45*activity + 0.35*rhythm + 0.20*longevity); archived -> min(score, 10)

## Adoption (not part of the score)
Stars 1726, forks 230 (observed 2026-08-28T04:05:28.155855+00:00)

## What it is
A PyTorch library providing differentiable stochastic differential equation (SDE) solvers with GPU support and efficient backpropagation via stochastic adjoint sensitivity analysis. It enables training neural SDEs and latent SDE models end-to-end with gradient-based optimization.

## Use cases
- solve stochastic differential equations on GPU in PyTorch
- train neural SDEs with backpropagation
- fit latent SDE models to time series data
- compute gradients through SDE solvers efficiently
- generate stochastic time series with a GAN trained SDE generator
- simulate Brownian-motion-driven dynamical systems

## When to choose
- you need differentiable SDE integration inside a PyTorch training loop
- you are building neural differential equation models with stochastic dynamics
- you need memory-efficient adjoint gradients for SDEs
- you want GPU-accelerated stochastic simulation of Ito or Stratonovich SDEs

## When to avoid
- you need deterministic ODE solvers only (use torchdiffeq instead)
- you work outside PyTorch, e.g. in JAX or TensorFlow
- you need production-hardened numerical SDE tooling rather than a research library
- your project requires frequent updates or active feature development

## Facets
- artifact type: library
- maturity: maintenance
- function: machine-learning, deep-learning, simulation, math
- domain: deep-learning, machine-learning, simulation
- platform: python, cross-platform
- tags: sde-solvers, pytorch, stochastic-differential-equations, neural-differential-equations, differentiable-programming, adjoint-sensitivity, brownian-motion, algorithms, gpu

## Member repositories
- google-research/torchsde (main) score 10

## Provenance
- Observed fields: from GitHub, fetched 2026-08-28T04:05:28.155855+00:00.
- Health v2: computed from the inputs above; adoption is never an input.
- Inferred fields (summary, facets, guidance): AI-extracted, prompt v1, taxonomy v1, on 2026-08-30T03:32:32.124748+00:00, confidence not recorded.
  - readme: https://github.com/google-research/torchsde (fetched 2026-08-28T04:05:28.155855+00:00, sha dbffd3d2ba91)
  - registry_pypi: https://pypi.org/pypi/torchsde/json (fetched 2026-08-29T11:09:04.588034+00:00, sha d61c252d4fec)
- Data as of 2026-08-30T08:39:29.467469+00:00.