TorchOpt is a high-performance optimizer library built upon PyTorch for easy implementation of functional optimization and gradient-based meta-learning. It consists of two main features:

  • TorchOpt provides functional optimizer which enables JAX-like composable functional optimizer for PyTorch. With TorchOpt, one can easily conduct neural network optimization in PyTorch with functional style optimizer, similar to Optax in JAX.

  • With the design of functional programming, TorchOpt provides efficient, flexible, and easy-to-implement differentiable optimizer for gradient-based meta-learning research. It largely reduces the efforts required to implement sophisticated meta-learning algorithms.



Please follow the instructions at to install PyTorch in your Python environment first. Then run the following command to install TorchOpt from PyPI:

pip install torchopt

You can also build shared libraries from source, use:

git clone
cd torchopt
pip3 install .

We provide a conda environment recipe to install the build toolchain such as cmake, g++, and nvcc:

git clone
cd torchopt

# You may need `CONDA_OVERRIDE_CUDA` if conda fails to detect the NVIDIA driver (e.g. in docker or WSL2)
CONDA_OVERRIDE_CUDA=11.7 conda env create --file conda-recipe-minimal.yaml

conda activate torchopt

Developer Documentation

The Team

TorchOpt is a work by


If you are having issues, please let us know by filing an issue on our issue tracker.




TorchOpt is licensed under the Apache 2.0 License.


If you find TorchOpt useful, please cite it in your publications.

  title   = {TorchOpt: An Efficient Library for Differentiable Optimization},
  author  = {Ren, Jie and Feng, Xidong and Liu, Bo and Pan, Xuehai and Fu, Yao and Mai, Luo and Yang, Yaodong},
  journal = {arXiv preprint arXiv:2211.06934},
  year    = {2022}

Indices and tables