Cheap and reliable Node.js hosting starts at $3/month, and $1/month static HTML hosting

Created with love in Canada, visit hostnodejs.com today

Feel like to post an Ad? Learn Details

All Projects → Lezcano → Geotorch

Lezcano / Geotorch

Licence: mit

Constrained optimization toolkit for PyTorch

Programming Languages

python

139335 projects - #7 most used programming language

Labels

pytorch

GeoTorch

A library for constrained optimization and manifold optimization for deep learning in PyTorch

Overview

GeoTorch provides a simple way to perform constrained optimization and optimization on manifolds in PyTorch. It is compatible out of the box with any optimizer, layer, and model implemented in PyTorch without having to reimplement the layers or optimizers and without any kind of boilerplate in the training code.

.. code:: python

import torch
import torch.nn as nn
import geotorch

class Net(nn.Module):
    def __init__(self):
        super(Model, self).__init__()
        self.linear = nn.Linear(64, 128)
        self.cnn = nn.Conv2d(16, 32, 3)
        # Make the linear layer into a low rank layer with rank at most 10
        geotorch.low_rank(self.linear, "weight", rank=10)
        # Also works on tensors. Makes every kernel orthogonal
        geotorch.orthogonal(self.cnn, "weight")

    def forward(self, x):
        # self.linear has rank at most 10 and every 3x3 kernel in the CNN is orthogonal

# Nothing fancy from here on. Use the model as you'd normally do.
model = Net()

# Use your optimizer of choice. Any optimizer works out of the box with any parametrization
optim = torch.optim.Adam(model.parameters(), lr=lr)

The constrained tensor is always initialized to a matrix sampled according to some standard distributions on each space. Even then, we may initialize the tensors manually by assigning to them:

.. code:: python

linear = nn.Linear(64, 64)
geotorch.orthogonal(linear, "weight")  # linear.weight is an orthogonal matrix sampled uniformly
linear.weight = torch.eye(64)          # linear.weight is the identity matrix

Constraints

The following constraints are implemented and may be used as in the example above:

|symmetric|_. Symmetric matrices
|skew_constr|_. Skew-symmetric matrices
|sphere_constr|_. Vectors of norm 1
|orthogonal|_. Matrices with orthogonal columns
|grassmannian|_. Skew-symmetric matrices
|almost_orthogonal|_. Matrices with singular values in the interval [1-λ, 1+λ]
|invertible|_. Invertible matrices with positive determinant
|low_rank|_. Matrices of rank at most r
|fixed_rank|_. Matrices of rank r
|positive_definite|_. Positive definite matrices
|positive_semidefinite|_. Positive semidefinite matrices
|positive_semidefinite_low_rank|_. Positive semidefinite matrices of rank at most r
|positive_semidefinite_fixed_rank|_. Positive semidefinite matrices of rank r

Each of these constraints have some extra parameters which can be used to tailor the behavior of each constraint to the problem in hand. For more on this, see the documentation.

These constraints are a fronted for the families of spaces listed below.

Supported Spaces

Each constraint in GeoTorch is implemented as a manifold. These give the user more flexibility on the options that they choose for each parametrization. All these support Riemannian Gradient Descent by default (more on this here_), but they also support optimization via any other PyTorch optimizer.

GeoTorch currently supports the following spaces:

|reals|_: Rⁿ. Unrestricted optimization
|sym|_: Vector space of symmetric matrices
|skew|_: Vector space of skew-symmetric matrices
|sphere|_: Sphere in Rⁿ. { x ∈ Rⁿ | ||x|| = 1 } ⊂ Rⁿ
|so|_: Manifold of n×n orthogonal matrices
|st|_: Manifold of n×k matrices with orthonormal columns
|almost|_: Manifold of n×k matrices with singular values in the interval [1-λ, 1+λ]
|grass|_: Manifold of k-dimensional subspaces in Rⁿ
|glp|_: Manifold of invertible n×n matrices with positive determinant
|low|_: Variety of n×k matrices of rank r or less
|fixed|_: Manifold of n×k matrices of rank r
|psd|_: Cone of n×n symmetric positive definite matrices``
|pssd|_: Cone of n×n symmetric positive semi-definite matrices
|pssdlow|_: Variety of n×n symmetric positive semi-definite matrices of rank r or less
|pssdfixed|_: Manifold of n×n symmetric positive semi-definite matrices of rank r
|product|_: Product of manifolds M₁ × ... × Mₖ

Every space of dimension (n, k) can be applied to tensors of shape (*, n, k), so we also get efficient parallel implementations of product spaces such as

ObliqueManifold(n,k): Matrix with unit length columns, Sⁿ⁻¹ × ...ᵏ⁾ × Sⁿ⁻¹

Using GeoTorch in your Code

The files in examples/copying_problem.py_ and examples/sequential_mnist.py_ serve as tutorials to see how to handle the initialization and usage of GeoTorch in some real code. They also show how to implement Riemannian Gradient Descent and some other tricks. For an introduction to how the library is actually implemented, see exmaples/parametrisations.ipynb_.

You may try GeoTorch installing it with

.. code:: bash

pip install git+https://github.com/Lezcano/geotorch/

GeoTorch is tested in Linux, Mac, and Windows environments for Python >= 3.6.

Sharing Weights, Parametrizations, and Normalizing Flows

If one wants to use a parametrized tensor in different places in their model, or uses one parametrized layer many times, for example in an RNN, it is recommended to wrap the forward pass as follows to avoid each parametrization to be computed many times:

.. code:: python

with geotorch.parametrize.cached():
    logits = model(input_)

Of course, this with statement may be used simply inside the forward function where the parametrized layer is used several times.

These ideas fall in the context of parametrized optimization, where one wraps a tensor X with a function f, and rather than using X, uses f(X). Particular examples of this idea are pruning, weight normalization, and spectral normalization among others. This repository implements a framework to approach this kind of problems. The framework is currently PR #33344_ in PyTorch. All the functionality of this PR is located in geotorch/parametrize.py_.

As every space in GeoTorch is, at its core, a map from a flat space into a manifold, the tools implemented here also serve as a building block in normalizing flows. Using a factorized space such as |low|_ it is direct to compute the determinant of the transformation it defines, as we have direct access to the singular values of the layer.

Bibliography

Please cite the following work if you found GeoTorch useful. This paper exposes a simplified mathematical explanation of part of the inner-workings of GeoTorch.

.. code:: bibtex

@inproceedings{lezcano2019trivializations,
    title = {Trivializations for gradient-based optimization on manifolds},
    author = {Lezcano-Casado, Mario},
    booktitle={Advances in Neural Information Processing Systems, NeurIPS},
    pages = {9154--9164},
    year = {2019},
}

.. |Build| image:: https://github.com/lezcano/geotorch/workflows/Build/badge.svg :target: https://github.com/lezcano/geotorch/workflows/Build/badge.svg :alt: Build .. |Docs| image:: https://readthedocs.org/projects/geotorch/badge/?version=latest :target: https://geotorch.readthedocs.io/en/latest/?badge=latest .. |Codecov| image:: https://codecov.io/gh/Lezcano/geotorch/branch/master/graph/badge.svg?token=1AKM2EQ7RT :target: https://codecov.io/gh/Lezcano/geotorch/branch/master/graph/badge.svg?token=1AKM2EQ7RT :alt: Code coverage .. |Codestyle Black| image:: https://img.shields.io/badge/code%20style-black-000000.svg :target: https://github.com/ambv/black :alt: Codestyle Black .. |License| image:: https://img.shields.io/badge/license-MIT-green.svg :target: https://github.com/Lezcano/geotorch/blob/master/LICENSE :alt: License

.. _here: https://github.com/Lezcano/geotorch/blob/master/examples/copying_problem.py#L16 .. _PR #33344: https://github.com/pytorch/pytorch/pull/33344 .. _geotorch/parametrize.py: https://github.com/Lezcano/geotorch/blob/master/geotorch/parametrize.py .. _examples/sequential_mnist.py: https://github.com/Lezcano/geotorch/blob/master/examples/sequential_mnist.py .. _examples/copying_problem.py: https://github.com/Lezcano/geotorch/blob/master/examples/copying_problem.py .. _examples/parametrisations.ipynb: https://github.com/Lezcano/geotorch/blob/master/examples/parametrisations.ipynb

Note that the project description data, including the texts, logos, images, and/or trademarks, for each open source project belongs to its rightful owner. If you wish to add or remove any projects, please contact us at [email protected].

Stars: ✭ 243

Visit Git Page 🔗Visit User Page 🔗Visit Issues Page (0) 🔗