#156156th of 242 in ML frameworks and libraries

botorch

Bayesian optimization library built on PyTorch and GPyTorch

Stars
3.7k
License
MIT
Last commit
Nov 2026
Last release
Jun 2026
Language
Jupyter Notebook
Stars, last 29 weeks+151
25 Apr 2026: 3,504 stars2 May 2026: 3,508 stars9 May 2026: 3,512 stars16 May 2026: 3,519 stars23 May 2026: 3,525 stars30 May 2026: 3,529 stars6 Jun 2026: 3,540 stars13 Jun 2026: 3,549 stars20 Jun 2026: 3,556 stars27 Jun 2026: 3,559 stars4 Jul 2026: 3,564 stars11 Jul 2026: 3,569 stars18 Jul 2026: 3,571 stars25 Jul 2026: 3,573 stars1 Aug 2026: 3,576 stars8 Aug 2026: 3,579 stars15 Aug 2026: 3,583 stars22 Aug 2026: 3,588 stars29 Aug 2026: 3,589 stars5 Sept 2026: 3,591 stars12 Sept 2026: 3,594 stars19 Sept 2026: 3,599 stars26 Sept 2026: 3,603 stars3 Oct 2026: 3,608 stars10 Oct 2026: 3,615 stars17 Oct 2026: 3,625 stars24 Oct 2026: 3,634 stars31 Oct 2026: 3,644 stars7 Nov 2026: 3,648 stars14 Nov 2026: 3,655 stars

3,504 on 25 Apr 20263,655 on 14 Nov 2026

From GitHub's daily star counts.

Checked on GitHub.

Overview

BoTorch is a Python library for composing Bayesian optimization loops from Gaussian process models, acquisition functions and optimizers. It uses PyTorch autodiff and supports Monte Carlo acquisition functions via the reparameterization trick, with GPyTorch models underneath. The README positions it as a low-level API for researchers and recommends Ax for end-users.

Who it is for: Researchers building Bayesian optimization algorithms in PyTorch

Strengths

  • Modular primitives: models, acquisition functions and optimizers compose separately
  • Monte Carlo acquisition functions via reparameterization, no restrictive model assumptions
  • First-class GPyTorch support, including multi-task GPs, deep kernel learning and deep GPs
  • Device-agnostic PyTorch code runs on GPUs with autodiff

Weaknesses

  • README describes it as beta and under active development
  • Aimed at researchers; non-researchers are directed to Ax instead
  • Requires Python 3.11+ plus gpytorch, linear_operator and pyro-ppl
  • Official Anaconda package stopped after 0.12; conda-forge only

What it needs

  • GPU optional
  • Needs PyTorch >= 2.0.1, gpytorch >= 1.14, linear_operator >= 0.6, pyro-ppl >= 1.8.4, scipy, multiple-dispatch
  • Models: Gaussian processes (GPyTorch), multi-task GPs, deep kernel learning, deep GPs

What changed

From the release notes. No flag means the note does not say so, not that the upgrade is safe.

Also in ML frameworks and libraries

See all 242
Also in ML frameworks and libraries
RankProjectScore
1transformersModel-definition library for pretrained text, vision, audio and multimodal models169.3k starsApache-2.094 out of 100
2ultralyticsPython library and CLI for training and running YOLO vision models63.5k starsAGPL-3.094 out of 100
3pytorch-image-modelsPyTorch library of image models, pretrained weights and training scripts37.3k starsApache-2.094 out of 100
#4pytorchPython tensor library with GPU support and tape-based autograd105.8k starscustom license93 out of 100
#5supervisionPython toolkit for loading, annotating and counting computer vision detections Live demo ↗ (opens in a new tab)51.7k starsMIT92 out of 100