A library for differentiable nonlinear optimization.

Meta Research

Last update: Dec 30, 2022

Related tags

Deep Learning theseus

Overview

Theseus

A library for differentiable nonlinear optimization built on PyTorch to support constructing various problems in robotics and vision as end-to-end differentiable architectures.

The current focus is on nonlinear least squares with support for sparsity, batching, GPU, and backward modes for unrolling, truncated and implicit, and sampling based differentiation. This library is in beta with expected full release in mid 2022.

Getting Started

Prerequisites
- We strongly recommend you install theseus in a venv or conda environment.
- Theseus requires torch installation. To install for your particular CPU/CUDA configuration, follow the instructions in the PyTorch website.

Installing

git clone https://github.com/facebookresearch/theseus.git && cd theseus
pip install -e .

Running unit tests
```
pytest theseus
```
See tutorials and examples to learn about the API and usage.

Additional Information

Use Github issues for questions, suggestions, and bugs.
See CONTRIBUTING if interested in helping out.
Theseus is being developed with the help of many contributors, see THANKS.

License

Theseus is MIT licensed. See the LICENSE for details.

Comments

Override Vector operators in Point2 and Point3
Motivation and Context

Overrides vector operations for Point2, Point3 #113

How Has This Been Tested

Types of changes

[ ] Docs change / refactoring / dependency upgrade

[ ] Bug fix (non-breaking change which fixes an issue)

[x] New feature (non-breaking change which adds functionality)

[ ] Breaking change (fix or feature that would cause existing functionality to change)

Checklist

[ ] My code follows the code style of this project.

[ ] My change requires a change to the documentation.

[ ] I have updated the documentation accordingly.

[ ] I have read the CONTRIBUTING document.

[ ] I have completed my CLA (see CONTRIBUTING)

[ ] I have added tests to cover my changes.

[ ] All new and existing tests passed.

CLA Signed
opened by jeffin07 15
error and errorSquaredNorm can optionally take in variable data
🚀 Feature

API improvement in Objective: error and errorSquaredNorm can optionally take in var_data which if passed would call update internally. Document the behavior that if var_data is passed this will update the objective.

Motivation

Facilitates usage for cases where only error needs to be queried (w/o running optimization or even updating the variables).

Pitch

Following ways to use this api after:

Get error on current internal values: call error without passing any var_data

Get error on new values w\ update to objective: call error and pass var_data

Get error on new values w\o update to objective: call error, pass var_data and True optional flag to not update objective

enhancement good first issue
opened by luisenp 13
Allow constant inputs to cost functions to be passed as floats
Motivation and Context

To close #38

How Has This Been Tested

Types of changes

[ ] Docs change / refactoring / dependency upgrade

[ ] Bug fix (non-breaking change which fixes an issue)

[x] New feature (non-breaking change which adds functionality)

[ ] Breaking change (fix or feature that would cause existing functionality to change)

Checklist

[x] My code follows the code style of this project.

[ ] My change requires a change to the documentation.

[ ] I have updated the documentation accordingly.

[x] I have read the CONTRIBUTING document.

[x] I have completed my CLA (see CONTRIBUTING)

[ ] I have added tests to cover my changes.

[ ] All new and existing tests passed.

CLA Signed refactor
opened by jeffin07 10
error and errorsquaredNorm optional data
Motivation and Context

API improvement in Objective: error and errorSquaredNorm can optionally take in var_data which if passed would call update internally. Document the behavior that if var_data is passed this will update the objective. #4

How Has This Been Tested

Testes using unit test and a custom example

Types of changes

[ ] Docs change / refactoring / dependency upgrade

[ ] Bug fix (non-breaking change which fixes an issue)

[x] New feature (non-breaking change which adds functionality)

[ ] Breaking change (fix or feature that would cause existing functionality to change)

Checklist

[ ] My code follows the code style of this project.

[ ] My change requires a change to the documentation.

[ ] I have updated the documentation accordingly.

[ ] I have read the CONTRIBUTING document.

[ ] I have completed my CLA (see CONTRIBUTING)

[x] I have added tests to cover my changes.

[ ] All new and existing tests passed.

enhancement CLA Signed
opened by jeffin07 10

Updating just `aux_vars` isn't sufficient to re-solve with some data changed

🐛 Bug

Using the quadratic fit example, I thought it would be reasonable to update the data in just aux_vars and re-solve, but it seems like there's a dependence on the global data_x.

Steps to Reproduce

I included a MWE below that outputs the incorrect solution in the middle here by just changing aux_inputs:

optimal a:  tensor([[1.0076]], grad_fn=<AddBackward0>)
== Only changing aux_vars["x"] (this should not be the same solution)
optimal a:  tensor([[1.0076]], grad_fn=<AddBackward0>)
== Globally updating data_x (this is the correct solution)
optimal a:  tensor([[0.0524]], grad_fn=<AddBackward0>)

Expected behavior

I was pretty confused at first when my code using wasn't working and didn't realize it was because of this. We should make updating aux_inputs sufficient to re-solve the problem, or if this is challenging we should consider 1) raising a warning/adding a check with aux_inputs doesn't match or 2) remove duplicated passing of aux_inputs when it doesn't do anything.

Code

#!/usr/bin/env python3

import torch
import theseus as th
import theseus.optimizer.nonlinear as thnl

import numpy as np
import numdifftools as nd

def generate_data(num_points=10, a=1., b=0.5, noise_factor=0.01):
    data_x = torch.rand((1, num_points))
    noise = torch.randn((1, num_points)) * noise_factor
    data_y = a * data_x.square() + b + noise
    return data_x, data_y

num_points = 10
data_x, data_y = generate_data(num_points)

x = th.Variable(data_x.requires_grad_(), name="x")
y = th.Variable(data_y.requires_grad_(), name="y")
a = th.Vector(1, name="a")
b = th.Vector(1, name="b")

def quad_error_fn(optim_vars, aux_vars):
    a, b = optim_vars
    x, y = aux_vars
    est = a.data * x.data.square() + b.data
    err = y.data - est
    return err

optim_vars = a, b
aux_vars = x, y
cost_function = th.AutoDiffCostFunction(
    optim_vars, quad_error_fn, num_points, aux_vars=aux_vars, name="quadratic_cost_fn"
)
objective = th.Objective()
objective.add(cost_function)
optimizer = th.GaussNewton(
    objective,
    max_iterations=15,
    step_size=0.5,
)

theseus_inputs = {
"a": 2 * torch.ones((1, 1)).requires_grad_(),
"b": torch.ones((1, 1)).requires_grad_()
}
aux_vars = {
"x": data_x,
"y": data_y,
}
theseus_optim = th.TheseusLayer(optimizer)
updated_inputs, info = theseus_optim.forward(
    theseus_inputs, aux_vars=aux_vars,
    track_best_solution=True, verbose=False,
    backward_mode=thnl.BackwardMode.FULL,
)
print('optimal a: ', updated_inputs['a'])

aux_vars = {
"x": data_x+10.,
"y": data_y,
}
updated_inputs, info = theseus_optim.forward(
    theseus_inputs, aux_vars=aux_vars,
    track_best_solution=True, verbose=False,
    backward_mode=thnl.BackwardMode.FULL,
)
print('== Only changing aux_vars["x"] (this should not be the same solution)')
print('optimal a: ', updated_inputs['a'])

data_x.data += 10.
aux_vars = {
"x": data_x,
"y": data_y,
}
updated_inputs, info = theseus_optim.forward(
    theseus_inputs, aux_vars=aux_vars,
    track_best_solution=True, verbose=False,
    backward_mode=thnl.BackwardMode.FULL,
)
print('== Globally updating data_x (this is the correct solution)')
print('optimal a: ', updated_inputs['a'])

documentation question refactor

opened by bamos 7

Refactored SE3.log_map_impl() to avoid in place operations

Fixes @exhaustin torch backward errors in this script.

The script still has other errors, where the final system is not positive definite so it cannot be solved with CholeskyDense. But unclear if this is related to Lie groups or something else.
bug CLA Signed

opened by luisenp 6

Installation in Docker : error in compilation of extlib/mat_mul.cu

Hi ! Would it be possible to have a Dockerfile with the right config ? I've tried many compatible versions of pytorch and CUDA, but I always get the same error when building theseus-ai.. In my last trial, I started from nvidia ngc pytorch container nvcr.io/nvidia/pytorch:21.06-py3, which is an Ubuntu 20.04 with cuda 11.3, python 3.8 and I reinstalled torch==1.10.1+cu113 version.

Here's the full error :

    /home/dir/theseus/theseus/extlib/mat_mult.cu(74): error: no instance of overloaded function "atomicAdd" matches the argument list
                argument types are: (double *, double)
    /home/dir/theseus/theseus/extlib/mat_mult.cu(239): error: no instance of overloaded function "atomicAdd" matches the argument list
                argument types are: (double *, double)
    2 errors detected in the compilation of "/home/dir/theseus/theseus/extlib/mat_mult.cu".
    ninja: build stopped: subcommand failed.
    Traceback (most recent call last):
      File "/opt/conda/lib/python3.8/site-packages/torch/utils/cpp_extension.py", line 1717, in _run_ninja_build
        subprocess.run(
      File "/opt/conda/lib/python3.8/subprocess.py", line 516, in run
        raise CalledProcessError(retcode, process.args,
    subprocess.CalledProcessError: Command '['ninja', '-v']' returned non-zero exit status 1.
    The above exception was the direct cause of the following exception:
    Traceback (most recent call last):
      File "<string>", line 1, in <module>
      File "/home/dir/theseus/setup.py", line 60, in <module>
        setuptools.setup(
      File "/opt/conda/lib/python3.8/site-packages/setuptools/__init__.py", line 163, in setup
        return distutils.core.setup(**attrs)
      File "/opt/conda/lib/python3.8/distutils/core.py", line 148, in setup
        dist.run_commands()
      File "/opt/conda/lib/python3.8/distutils/dist.py", line 966, in run_commands
        self.run_command(cmd)
      File "/opt/conda/lib/python3.8/distutils/dist.py", line 985, in run_command
        cmd_obj.run()
      File "/opt/conda/lib/python3.8/site-packages/setuptools/command/develop.py", line 38, in run
        self.install_for_development()
      File "/opt/conda/lib/python3.8/site-packages/setuptools/command/develop.py", line 140, in install_for_development
        self.run_command('build_ext')
      File "/opt/conda/lib/python3.8/distutils/cmd.py", line 313, in run_command
        self.distribution.run_command(command)
      File "/opt/conda/lib/python3.8/distutils/dist.py", line 985, in run_command
        cmd_obj.run()
      File "/opt/conda/lib/python3.8/site-packages/setuptools/command/build_ext.py", line 87, in run
        _build_ext.run(self)
      File "/opt/conda/lib/python3.8/site-packages/Cython/Distutils/old_build_ext.py", line 186, in run
        _build_ext.build_ext.run(self)
      File "/opt/conda/lib/python3.8/distutils/command/build_ext.py", line 340, in run
        self.build_extensions()
      File "/opt/conda/lib/python3.8/site-packages/torch/utils/cpp_extension.py", line 735, in build_extensions
        build_ext.build_extensions(self)
      File "/opt/conda/lib/python3.8/site-packages/Cython/Distutils/old_build_ext.py", line 194, in build_extensions
        self.build_extension(ext)
      File "/opt/conda/lib/python3.8/site-packages/setuptools/command/build_ext.py", line 208, in build_extension
        _build_ext.build_extension(self, ext)
      File "/opt/conda/lib/python3.8/distutils/command/build_ext.py", line 528, in build_extension
        objects = self.compiler.compile(sources,
      File "/opt/conda/lib/python3.8/site-packages/torch/utils/cpp_extension.py", line 556, in unix_wrap_ninja_compile
        _write_ninja_file_and_compile_objects(
      File "/opt/conda/lib/python3.8/site-packages/torch/utils/cpp_extension.py", line 1399, in _write_ninja_file_and_compile_objects
        _run_ninja_build(
      File "/opt/conda/lib/python3.8/site-packages/torch/utils/cpp_extension.py", line 1733, in _run_ninja_build
        raise RuntimeError(message) from e
    RuntimeError: Error compiling objects for extension
    ----------------------------------------
ERROR: Command errored out with exit status 1: /opt/conda/bin/python3.8 -c 'import io, os, sys, setuptools, tokenize; sys.argv[0] = '"'"'/home/dir/theseus/setup.py'"'"'; __file__='"'"'/home/dir/theseus/setup.py'"'"';f = getattr(tokenize, '"'"'open'"'"', open)(__file__) if os.path.exists(__file__) else io.StringIO('"'"'from setuptools import setup; setup()'"'"');code = f.read().replace('"'"'\r\n'"'"', '"'"'\n'"'"');f.close();exec(compile(code, __file__, '"'"'exec'"'"'))' develop --no-deps Check the logs for full command output.

opened by fmagera 5

Homography example with functorch
Motivation and Context

How Has This Been Tested

Types of changes

[ ] Docs change / refactoring / dependency upgrade

[ ] Bug fix (non-breaking change which fixes an issue)

[ ] New feature (non-breaking change which adds functionality)

[ ] Breaking change (fix or feature that would cause existing functionality to change)

Checklist

[ ] My code follows the code style of this project.

[ ] My change requires a change to the documentation.

[ ] I have updated the documentation accordingly.

[ ] I have read the CONTRIBUTING document.

[ ] I have completed my CLA (see CONTRIBUTING)

[ ] I have added tests to cover my changes.

[ ] All new and existing tests passed.

CLA Signed
opened by fantaosha 5
Add robust cost function
Motivation and Context

How Has This Been Tested

Types of changes

[ ] Docs change / refactoring / dependency upgrade

[ ] Bug fix (non-breaking change which fixes an issue)

[ ] New feature (non-breaking change which adds functionality)

[ ] Breaking change (fix or feature that would cause existing functionality to change)

Checklist

[ ] My code follows the code style of this project.

[ ] My change requires a change to the documentation.

[ ] I have updated the documentation accordingly.

[ ] I have read the CONTRIBUTING document.

[ ] I have completed my CLA (see CONTRIBUTING)

[ ] I have added tests to cover my changes.

[ ] All new and existing tests passed.

CLA Signed
opened by fantaosha 5
Add ManifoldGaussian class for messages in belief propagation
Motivation and Context

It would be useful to have an optional covariance / precision matrix as part of the Manifold class as Gaussian Belief Propagation involves sending Gaussian distributions over the Manifold variables. Currently I'm using a wrapper Gaussian class but it could be more widely useful to have covariance / precision matrix as an attribute of the Manifold class?

How Has This Been Tested

Types of changes

[ ] Docs change / refactoring / dependency upgrade

[ ] Bug fix (non-breaking change which fixes an issue)

[ ] New feature (non-breaking change which adds functionality)

[ ] Breaking change (fix or feature that would cause existing functionality to change)

Checklist

[ ] My code follows the code style of this project.

[ ] My change requires a change to the documentation.

[ ] I have updated the documentation accordingly.

[ ] I have read the CONTRIBUTING document.

[ ] I have completed my CLA (see CONTRIBUTING)

[ ] I have added tests to cover my changes.

[ ] All new and existing tests passed.

enhancement CLA Signed
opened by joeaortiz 5
Override `Vector` operators in `Point2` and `Point3` with the correct return type
🚀 Feature

Something like

class Point2(Vector): ... def __add__(self, other: Vector) -> "Point2": return cast(Point2, super().__add__(other))

Motivation

Eliminates unnecessary casting when using typing.

Alternatives

There might be some way for mypy to do the correct thing that wouldn't require overriding these methods.

Additional context
enhancement good first issue
opened by luisenp 5
Add differentiable forward kinematics
Motivation and Context

How Has This Been Tested

Types of changes

[ ] Docs change / refactoring / dependency upgrade

[ ] Bug fix (non-breaking change which fixes an issue)

[ ] New feature (non-breaking change which adds functionality)

[ ] Breaking change (fix or feature that would cause existing functionality to change)

Checklist

[ ] My code follows the code style of this project.

[ ] My change requires a change to the documentation.

[ ] I have updated the documentation accordingly.

[ ] I have read the CONTRIBUTING document.

[ ] I have completed my CLA (see CONTRIBUTING)

[ ] I have added tests to cover my changes.

[ ] All new and existing tests passed.

enhancement CLA Signed
opened by fantaosha 0
CUDA kernel for differentiable sparse matrix vector product

Backward pass can be made more efficient in GPU if we write a custom CUDA kernel for it, but this should be reasonable enough for now.

_Originally posted by @luisenp in https://github.com/facebookresearch/theseus/pull/392

opened by mhmukadam 0
Add robot model
Motivation and Context

How Has This Been Tested

Types of changes

[ ] Docs change / refactoring / dependency upgrade

[ ] Bug fix (non-breaking change which fixes an issue)

[ ] New feature (non-breaking change which adds functionality)

[ ] Breaking change (fix or feature that would cause existing functionality to change)

Checklist

[ ] My code follows the code style of this project.

[ ] My change requires a change to the documentation.

[ ] I have updated the documentation accordingly.

[ ] I have read the CONTRIBUTING document.

[ ] I have completed my CLA (see CONTRIBUTING)

[ ] I have added tests to cover my changes.

[ ] All new and existing tests passed.

CLA Signed
opened by fantaosha 0
Add prismatic joint
Motivation and Context

How Has This Been Tested

Types of changes

[ ] Docs change / refactoring / dependency upgrade

[ ] Bug fix (non-breaking change which fixes an issue)

[ ] New feature (non-breaking change which adds functionality)

[ ] Breaking change (fix or feature that would cause existing functionality to change)

Checklist

[ ] My code follows the code style of this project.

[ ] My change requires a change to the documentation.

[ ] I have updated the documentation accordingly.

[ ] I have read the CONTRIBUTING document.

[ ] I have completed my CLA (see CONTRIBUTING)

[ ] I have added tests to cover my changes.

[ ] All new and existing tests passed.

enhancement CLA Signed
opened by fantaosha 0
Add revolute joint
Motivation and Context

How Has This Been Tested

Types of changes

[ ] Docs change / refactoring / dependency upgrade

[ ] Bug fix (non-breaking change which fixes an issue)

[ ] New feature (non-breaking change which adds functionality)

[ ] Breaking change (fix or feature that would cause existing functionality to change)

Checklist

[ ] My code follows the code style of this project.

[ ] My change requires a change to the documentation.

[ ] I have updated the documentation accordingly.

[ ] I have read the CONTRIBUTING document.

[ ] I have completed my CLA (see CONTRIBUTING)

[ ] I have added tests to cover my changes.

[ ] All new and existing tests passed.

enhancement CLA Signed
opened by fantaosha 0
Add se3.log()
Motivation and Context

How Has This Been Tested

Types of changes

[ ] Docs change / refactoring / dependency upgrade

[ ] Bug fix (non-breaking change which fixes an issue)

[ ] New feature (non-breaking change which adds functionality)

[ ] Breaking change (fix or feature that would cause existing functionality to change)

Checklist

[ ] My code follows the code style of this project.

[ ] My change requires a change to the documentation.

[ ] I have updated the documentation accordingly.

[ ] I have read the CONTRIBUTING document.

[ ] I have completed my CLA (see CONTRIBUTING)

[ ] I have added tests to cover my changes.

[ ] All new and existing tests passed.

enhancement CLA Signed
opened by fantaosha 0

Releases(0.1.3)

0.1.3(Nov 9, 2022)
Major Updates

Adaptive damping for Levenberg-Marquardt by @luisenp in https://github.com/facebookresearch/theseus/pull/328

Moved all unit tests to a separate folder by @luisenp in https://github.com/facebookresearch/theseus/pull/352

Other Changes

Removed manual cmake install for CPU tests. by @luisenp in https://github.com/facebookresearch/theseus/pull/338

Fixed vmap related bug breaking homography ex. with sparse solvers. by @luisenp in https://github.com/facebookresearch/theseus/pull/337

Small vectorization improvements by @luisenp in https://github.com/facebookresearch/theseus/pull/336

Change CI to separately handle torch >= 1.13 by @luisenp in https://github.com/facebookresearch/theseus/pull/345

Fixed quaternion bug at pi by @fantaosha in https://github.com/facebookresearch/theseus/pull/344

Expose Lie Groups checks at root level by @luisenp in https://github.com/facebookresearch/theseus/pull/335

Added option for making LieGroup checks silent. by @luisenp in https://github.com/facebookresearch/theseus/pull/351

Added a few other CUDA versions to build script. by @luisenp in https://github.com/facebookresearch/theseus/pull/349

Set vectorization off by default when using optimizers w/o TheseusLayer by @luisenp in https://github.com/facebookresearch/theseus/pull/350

Some more cleanup before 0.1.3 by @luisenp in https://github.com/facebookresearch/theseus/pull/353

Add tests for wheels in CI by @luisenp in https://github.com/facebookresearch/theseus/pull/354

#355 Add device parameter to UrdfRobotModel by @thomasweng15 in https://github.com/facebookresearch/theseus/pull/356

Added th.device to represent both str and torch.device. by @luisenp in https://github.com/facebookresearch/theseus/pull/357

Update to 0.1.3 by @luisenp in https://github.com/facebookresearch/theseus/pull/358

New Contributors

@thomasweng15 made their first contribution in https://github.com/facebookresearch/theseus/pull/356

Full Changelog: https://github.com/facebookresearch/theseus/compare/0.1.2...0.1.3
Source code(tar.gz)
Source code(zip)
0.1.2(Oct 20, 2022)
Major Updates

Add support for BaSpaCho sparse solver by @maurimo in https://github.com/facebookresearch/theseus/pull/324

Set vmap as the default autograd mode for autodiff cost function. by @luisenp in https://github.com/facebookresearch/theseus/pull/313

Other Changes

Changed homography example to allow benchmarking only cost computation by @luisenp in https://github.com/facebookresearch/theseus/pull/311

Run isort on all files by @luisenp in https://github.com/facebookresearch/theseus/pull/312

Added usort:skip tags. by @luisenp in https://github.com/facebookresearch/theseus/pull/314

Fixed checkout tag syntax in build script. by @luisenp in https://github.com/facebookresearch/theseus/pull/315

Removed redundant directory in homography gif save. by @luisenp in https://github.com/facebookresearch/theseus/pull/316

Fixing simple example by @Gralerfics in https://github.com/facebookresearch/theseus/pull/320

AutodiffCostFunction now expands tensors with batch size 1 before running vmap by @luisenp in https://github.com/facebookresearch/theseus/pull/327

Deprecated FULL backward mode (now UNROLL). by @luisenp in https://github.com/facebookresearch/theseus/pull/332

Using better names for CHOLMOD solver python files. by @luisenp in https://github.com/facebookresearch/theseus/pull/333

New Contributors

@Gralerfics made their first contribution in https://github.com/facebookresearch/theseus/pull/320

Full Changelog: https://github.com/facebookresearch/theseus/compare/0.1.1...0.1.2
Source code(tar.gz)
Source code(zip)
0.1.1(Sep 28, 2022)
Highlights

Added pip install theseus-ai instructions. by @luisenp in https://github.com/facebookresearch/theseus/pull/276

Add functorch support for AutoDiffCostFunction by @fantaosha in https://github.com/facebookresearch/theseus/pull/268

Profile AutoDiffCostFunction and refactor the homography example by @fantaosha in https://github.com/facebookresearch/theseus/pull/296

What's Changed

update pose graph data link by @fantaosha in https://github.com/facebookresearch/theseus/pull/256

[homography] Use kornia lib properly for perspective transform by @ddetone in https://github.com/facebookresearch/theseus/pull/258

Small Update to 00_introduction.ipynb by @NeilPandya in https://github.com/facebookresearch/theseus/pull/259

Changed SDF constructor to accept more convenient data types. by @luisenp in https://github.com/facebookresearch/theseus/pull/260

Fixed small bugs in MotionPlanner class by @luisenp in https://github.com/facebookresearch/theseus/pull/261

Added option to visualize SDF to trajectory visualization function by @luisenp in https://github.com/facebookresearch/theseus/pull/263

update readme by @mhmukadam in https://github.com/facebookresearch/theseus/pull/264

Added MotionPlanner.forward() method. by @luisenp in https://github.com/facebookresearch/theseus/pull/267

Small bug fixes and tweaks to generate_trajectory_figs. by @luisenp in https://github.com/facebookresearch/theseus/pull/271

Added a script for building wheels on a new docker image. by @luisenp in https://github.com/facebookresearch/theseus/pull/257

Bugfix: homography estimation - create data folder before downloading data by @luizgh in https://github.com/facebookresearch/theseus/pull/275

Added pip install theseus-ai instructions. by @luisenp in https://github.com/facebookresearch/theseus/pull/276

Refactored MotionPlanner so that objective can be passed separately. by @luisenp in https://github.com/facebookresearch/theseus/pull/277

add numel() to Manifold and Lie groups by @fantaosha in https://github.com/facebookresearch/theseus/pull/280

Add support for SE2 poses in Collision2D by @luisenp in https://github.com/facebookresearch/theseus/pull/278

Probabilistically correct SO(3) sampling by @brentyi in https://github.com/facebookresearch/theseus/pull/286

Refactor SO3 and SE3 to be consistent with functorch by @fantaosha in https://github.com/facebookresearch/theseus/pull/266

Add SE2 support in MotionPlanner by @luisenp in https://github.com/facebookresearch/theseus/pull/282

Fixed bug in visualization of SE2 motion plans. by @luisenp in https://github.com/facebookresearch/theseus/pull/293

Add functorch support for AutoDiffCostFunction by @fantaosha in https://github.com/facebookresearch/theseus/pull/268

Changed requirements so that main.txt only includes essential dependencies by @luisenp in https://github.com/facebookresearch/theseus/pull/294

Add to_quaternion, rotation, translation and convention comment by @fantaosha in https://github.com/facebookresearch/theseus/pull/295

Added th.as_variable() function to simplify creating new variables. by @luisenp in https://github.com/facebookresearch/theseus/pull/299

Added an optional end-of-step callback to NonlinearOptimizer.optimize(). by @luisenp in https://github.com/facebookresearch/theseus/pull/297

Add AutogradMode to AutoDiffCostFunction by @fantaosha in https://github.com/facebookresearch/theseus/pull/300

Profile AutoDiffCostFunction and refactor the homography example by @fantaosha in https://github.com/facebookresearch/theseus/pull/296

Changed unit tests so that the batch sizes to tests are defined in a central import by @luisenp in https://github.com/facebookresearch/theseus/pull/298

enhance the efficiency of Objectve.add() by @Christopher6488 in https://github.com/facebookresearch/theseus/pull/303

Added missing end newlines by @luisenp in https://github.com/facebookresearch/theseus/pull/307

Rename BackwardMode.FULL --> UNROLL and simplify backward mode config by @luisenp in https://github.com/facebookresearch/theseus/pull/305

Simplified autograd mode specification. by @luisenp in https://github.com/facebookresearch/theseus/pull/306

Clean up test_theseus_layer by @luisenp in https://github.com/facebookresearch/theseus/pull/308

update readme and bump version by @mhmukadam in https://github.com/facebookresearch/theseus/pull/309

New Contributors

@NeilPandya made their first contribution in https://github.com/facebookresearch/theseus/pull/259

@luizgh made their first contribution in https://github.com/facebookresearch/theseus/pull/275

@brentyi made their first contribution in https://github.com/facebookresearch/theseus/pull/286

@Christopher6488 made their first contribution in https://github.com/facebookresearch/theseus/pull/303

Full Changelog: https://github.com/facebookresearch/theseus/compare/0.1.0...0.1.1
Source code(tar.gz)
Source code(zip)
0.1.0(Jul 20, 2022)
What's Changed

Add SO3 support by @luisenp in https://github.com/facebookresearch/theseus/pull/46

Taoshaf.add so3 class by @fantaosha in https://github.com/facebookresearch/theseus/pull/65

Add SO3 rotate and unrotate by @fantaosha in https://github.com/facebookresearch/theseus/pull/57

add so3._hat_matrix_check() by @fantaosha in https://github.com/facebookresearch/theseus/pull/59

Encapsulated data loading functions in tactile pushing example into a new class by @luisenp in https://github.com/facebookresearch/theseus/pull/51

Added a TactilePoseEstimator class to easily create TheseusLayer for tactile pushing by @luisenp in https://github.com/facebookresearch/theseus/pull/52

Refactor tactile pushing model interface by @luisenp in https://github.com/facebookresearch/theseus/pull/55

Minor fixes 02/01/2022 by @luisenp in https://github.com/facebookresearch/theseus/pull/66

add adjoint, hat and vee for SE3 by @fantaosha in https://github.com/facebookresearch/theseus/pull/68

Add SE3.exp_map() and SE3.log_map() by @fantaosha in https://github.com/facebookresearch/theseus/pull/71

Add SE3.compose() by @fantaosha in https://github.com/facebookresearch/theseus/pull/72

add SE3.transform_from and SE3.transform_to by @fantaosha in https://github.com/facebookresearch/theseus/pull/80

Updated to CircleCI's next-gen images. by @luisenp in https://github.com/facebookresearch/theseus/pull/89

Updated README with libsuitesparse installation instructions. by @luisenp in https://github.com/facebookresearch/theseus/pull/90

Added kwarg to NonlinearOptimizer.optimizer() for tracking error history by @luisenp in https://github.com/facebookresearch/theseus/pull/82

Merge infos results for truncated backward modes by @luisenp in https://github.com/facebookresearch/theseus/pull/83

Fix Issue 88 by @maurimo in https://github.com/facebookresearch/theseus/pull/97

Fixed bug in Variable.update() that was breaking torch graph... by @luisenp in https://github.com/facebookresearch/theseus/pull/96

Forced Gauss-Newton step for last iterations of truncated backward. by @luisenp in https://github.com/facebookresearch/theseus/pull/81

Add automatic differentiation on the Lie group tangent space by @fantaosha in https://github.com/facebookresearch/theseus/pull/74

Add rand() to LieGroup by @fantaosha in https://github.com/facebookresearch/theseus/pull/95

Fix SO2 rotate and unrotate jacobian by @fantaosha in https://github.com/facebookresearch/theseus/pull/58

Add projection for sparse Jacobian matrices by @fantaosha in https://github.com/facebookresearch/theseus/pull/98

Add LieGroup Support for AutoDiffFunction by @fantaosha in https://github.com/facebookresearch/theseus/pull/99

Add SE2.transform from() and fix shape bugs in SE2.transform_to() by @fantaosha in https://github.com/facebookresearch/theseus/pull/103

Switch SE3.transform_from and SE3.transform_to by @fantaosha in https://github.com/facebookresearch/theseus/pull/104

Enabled back ellipsoidal damping in LM with linear solvers support checks by @luisenp in https://github.com/facebookresearch/theseus/pull/87

Changed name of pytest mark for CUDA extensions. by @luisenp in https://github.com/facebookresearch/theseus/pull/102

Fix backpropagation bugs in SO3 and SE3 log_map by @fantaosha in https://github.com/facebookresearch/theseus/pull/109

Add analytical jacobians for LieGroup.exp_map by @fantaosha in https://github.com/facebookresearch/theseus/pull/110

error and errorsquaredNorm optional data by @jeffin07 in https://github.com/facebookresearch/theseus/pull/105

Add analytical derivatives for LieGroup.log_map() by @fantaosha in https://github.com/facebookresearch/theseus/pull/114

Refactor SO3.to_quaternion to fix backward bugs and improve the accuracy around pi by @fantaosha in https://github.com/facebookresearch/theseus/pull/116

Change RobotModel.forward_kinematics() interface by @luisenp in https://github.com/facebookresearch/theseus/pull/94

Added error handling for missing matplotlib and omegaconf installation when using theg by @luisenp in https://github.com/facebookresearch/theseus/pull/119

Add jacobians argument to exp_map and log_map functions by @joeaortiz in https://github.com/facebookresearch/theseus/pull/122

Add pose graph optimization example by @fantaosha in https://github.com/facebookresearch/theseus/pull/118

Added 'secret' option to keep constant step size when using truncated… by @luisenp in https://github.com/facebookresearch/theseus/pull/130

Add jacobians for LieGroup.local() by @fantaosha in https://github.com/facebookresearch/theseus/pull/129

Added method to update SE2 from x_y_theta. by @luisenp in https://github.com/facebookresearch/theseus/pull/131

A complete version of Bundle Adjusment by @luisenp in https://github.com/facebookresearch/theseus/pull/117

Fixed jacobians for Between and VariableDifference by @fantaosha in https://github.com/facebookresearch/theseus/pull/133

Added batching support to tactile pushing example. by @luisenp in https://github.com/facebookresearch/theseus/pull/132

Changed tactile pose example optim var initialization to use start pose for all vars by @luisenp in https://github.com/facebookresearch/theseus/pull/137

Override Vector operators in Point2 and Point3 by @jeffin07 in https://github.com/facebookresearch/theseus/pull/124

Merge Between and VariableDiff with RelativePoseError and PosePriorError by @fantaosha in https://github.com/facebookresearch/theseus/pull/136

Fix a bug in Objective.copy() by @fantaosha in https://github.com/facebookresearch/theseus/pull/139

Added option to force max iterations for TactilePoseEstimator. by @luisenp in https://github.com/facebookresearch/theseus/pull/141

black version bump by @jeffin07 in https://github.com/facebookresearch/theseus/pull/144

Added code to split tactile pushing trajectories data into train/val by @luisenp in https://github.com/facebookresearch/theseus/pull/143

Removed RobotModel.dim() by @luisenp in https://github.com/facebookresearch/theseus/pull/156

[bug-fix] Fixed wrong data shape initialization for GPCostWeight.dt by @luisenp in https://github.com/facebookresearch/theseus/pull/157

Made AutoDiffCostFunction._tmp_optim_vars copies of original by @luisenp in https://github.com/facebookresearch/theseus/pull/155

Add forward kinematics using an URDF to theseus.embodied.kinematics. by @exhaustin in https://github.com/facebookresearch/theseus/pull/84

Fixed dtype error in se3.py that came up in unit tests by @joeaortiz in https://github.com/facebookresearch/theseus/pull/158

Add-ons for backward experiments on Tactile Pose Estimation by @luisenp in https://github.com/facebookresearch/theseus/pull/164

Change unit tests to avoid making mypy a main requirement. by @luisenp in https://github.com/facebookresearch/theseus/pull/168

Update readme and contrib by @luisenp in https://github.com/facebookresearch/theseus/pull/169

Add ManifoldGaussian class for messages in belief propagation by @joeaortiz in https://github.com/facebookresearch/theseus/pull/121

More efficient implementation of forward kinematics by @exhaustin in https://github.com/facebookresearch/theseus/pull/175

Changing setup virtualenv command. by @luisenp in https://github.com/facebookresearch/theseus/pull/178

Updated SDF object in collision cost functions whenever an aux var is updated by @luisenp in https://github.com/facebookresearch/theseus/pull/177

Fixed device bug that occurred when merging info in TRUNCATED backward modes by @luisenp in https://github.com/facebookresearch/theseus/pull/181

Allow constant inputs to cost functions to be passed as floats by @jeffin07 in https://github.com/facebookresearch/theseus/pull/150

Minor changes to core test code. by @luisenp in https://github.com/facebookresearch/theseus/pull/197

adding logo by @mhmukadam in https://github.com/facebookresearch/theseus/pull/200

Added aliases for Difference and Between by @luisenp in https://github.com/facebookresearch/theseus/pull/199

Fixed infinite recursion in GPMotionModel.copy() by @luisenp in https://github.com/facebookresearch/theseus/pull/201

Fix bug in diagonal cost weight by @luisenp in https://github.com/facebookresearch/theseus/pull/203

Added a check in TheseusFunction that enforces copy() also copies the variables by @luisenp in https://github.com/facebookresearch/theseus/pull/202

Fixed bug in Objective.error() that was updating data unncessarily. by @luisenp in https://github.com/facebookresearch/theseus/pull/204

DLM gradients by @rtqichen in https://github.com/facebookresearch/theseus/pull/161

Setup now uses torch for checking CUDA availability, and CI runs py3.9 tests by @luisenp in https://github.com/facebookresearch/theseus/pull/206

Update README to specify python versions and CUDA during install step + by @cpaxton in https://github.com/facebookresearch/theseus/pull/207

Vectorization refactor by @luisenp in https://github.com/facebookresearch/theseus/pull/205

Implement Robust Cost Function by @luisenp in https://github.com/facebookresearch/theseus/pull/148

Added option for auto resetting LUCudaSparseSolver if the batch size needs to change by @luisenp in https://github.com/facebookresearch/theseus/pull/212

Address comments on RobustCostFunction implementation by @luisenp in https://github.com/facebookresearch/theseus/pull/213

Moved the method that retracts all variables with a given delta to Objective by @luisenp in https://github.com/facebookresearch/theseus/pull/214

Fixed flaky unit test for Collision2D jacobians. by @luisenp in https://github.com/facebookresearch/theseus/pull/216

Bundle Adjustment using RobustCostFunction by @luisenp in https://github.com/facebookresearch/theseus/pull/149

Fixed the cross product bug for SE3.exp_map and SE3.log_map by @fantaosha in https://github.com/facebookresearch/theseus/pull/217

Vectorize optimization variables retraction step by @luisenp in https://github.com/facebookresearch/theseus/pull/215

Moved Vectorize(objective) to the Optimizer class. by @luisenp in https://github.com/facebookresearch/theseus/pull/218

Added on-demand vectorization and also vectorized Objective.error(). by @luisenp in https://github.com/facebookresearch/theseus/pull/221

Vectorize PGO by @luisenp in https://github.com/facebookresearch/theseus/pull/211

Destroy cusolver context in CusolverLUSolver destructor by @luisenp in https://github.com/facebookresearch/theseus/pull/222

Jacobian computation using loop by @fantaosha in https://github.com/facebookresearch/theseus/pull/225

Vectorization handles singleton costs by @luisenp in https://github.com/facebookresearch/theseus/pull/226

Added close-formed jacobian for DLM perturbation cost. by @luisenp in https://github.com/facebookresearch/theseus/pull/224

SE2/SE3/SO3 - consolidate EPS, add dtype-conditioned EPS and add float32 unit tests by @luisenp in https://github.com/facebookresearch/theseus/pull/220

Add normalization to Lie group by @fantaosha in https://github.com/facebookresearch/theseus/pull/227

Add normalization to Lie group constructor by @fantaosha in https://github.com/facebookresearch/theseus/pull/228

Benchmarking PGO on main branch by @fantaosha in https://github.com/facebookresearch/theseus/pull/233

Renamed Variable.data as Variable.tensor by @luisenp in https://github.com/facebookresearch/theseus/pull/229

More data -> tensor renaming by @luisenp in https://github.com/facebookresearch/theseus/pull/230

Added state history tracking by @luisenp in https://github.com/facebookresearch/theseus/pull/234

Unified all cost functions so that cost weight is the last non-default argument by @luisenp in https://github.com/facebookresearch/theseus/pull/235

Ensure that CHOLMOD python interface casts to 64-bit precision by @luisenp in https://github.com/facebookresearch/theseus/pull/238

Add sphinx and readthedocs configuration by @luisenp in https://github.com/facebookresearch/theseus/pull/237

Fixed bug in state history for matrix data tensors. by @luisenp in https://github.com/facebookresearch/theseus/pull/240

Added isort for examples folder by @luisenp in https://github.com/facebookresearch/theseus/pull/243

Avoid batching SDF data, since it's shared by all trajectories. by @luisenp in https://github.com/facebookresearch/theseus/pull/246

Fixed device bug in DLM perturbation's jacobians by @luisenp in https://github.com/facebookresearch/theseus/pull/247

Benchmark PGO on the main branch by @fantaosha in https://github.com/facebookresearch/theseus/pull/244

Added checks to enforce 32- or 64-bit dtype by @luisenp in https://github.com/facebookresearch/theseus/pull/245

Update readme by @mhmukadam in https://github.com/facebookresearch/theseus/pull/251

Added MANIFEST.in and changed project name to theseus-ai. by @luisenp in https://github.com/facebookresearch/theseus/pull/252

Adding simple example by @mhmukadam in https://github.com/facebookresearch/theseus/pull/253

Added option to clear cuda cache when vectorization cache is cleared. by @luisenp in https://github.com/facebookresearch/theseus/pull/249

Added evaluation directory by @luisenp in https://github.com/facebookresearch/theseus/pull/241

New Contributors

@jeffin07 made their first contribution in https://github.com/facebookresearch/theseus/pull/105

@joeaortiz made their first contribution in https://github.com/facebookresearch/theseus/pull/122

@exhaustin made their first contribution in https://github.com/facebookresearch/theseus/pull/84

@rtqichen made their first contribution in https://github.com/facebookresearch/theseus/pull/161

@cpaxton made their first contribution in https://github.com/facebookresearch/theseus/pull/207

Full Changelog: https://github.com/facebookresearch/theseus/compare/0.1.0-b.2...0.1.0
Source code(tar.gz)
Source code(zip)
0.1.0-b.2(Feb 1, 2022)
Major Additions

Initial implicit/truncated backward modes by @bamos in https://github.com/facebookresearch/theseus/pull/29

Adds support for energy based learning with NLL loss (LEO) by @psodhi in https://github.com/facebookresearch/theseus/pull/30

cusolver based batched LU solver by @maurimo in https://github.com/facebookresearch/theseus/pull/22

CUDA batch matrix multiplication and ops by @maurimo in https://github.com/facebookresearch/theseus/pull/23

CUDA-based solver class and autograd function by @maurimo in https://github.com/facebookresearch/theseus/pull/24

What Else Changed

Added clearer explanation at the end of Tutorial 0 and fixed doc typos by @luisenp in https://github.com/facebookresearch/theseus/pull/2

Default SE2/SO2 is zero element rather than torch empty. by @luisenp in https://github.com/facebookresearch/theseus/pull/3

Add plots to tutorials by @bamos in https://github.com/facebookresearch/theseus/pull/25

update text in Tutorial 2 per issue #27 by @vshobha in https://github.com/facebookresearch/theseus/pull/31

Update contrib and add gitattributes by @mhmukadam in https://github.com/facebookresearch/theseus/pull/33

update continuous integration by @maurimo in https://github.com/facebookresearch/theseus/pull/21

Changed TheseusLayer.forward() to receive optimizer_kwargs as a single dict by @luisenp in https://github.com/facebookresearch/theseus/pull/45

[hotfix] fix lint issues by @maurimo in https://github.com/facebookresearch/theseus/pull/54

Update version by @mhmukadam in https://github.com/facebookresearch/theseus/pull/63

New Contributors

@luisenp made their first contribution in https://github.com/facebookresearch/theseus/pull/2

@bamos made their first contribution in https://github.com/facebookresearch/theseus/pull/25

@vshobha made their first contribution in https://github.com/facebookresearch/theseus/pull/31

@maurimo made their first contribution in https://github.com/facebookresearch/theseus/pull/21

@mhmukadam made their first contribution in https://github.com/facebookresearch/theseus/pull/33

@psodhi made their first contribution in https://github.com/facebookresearch/theseus/pull/30

Full Changelog: https://github.com/facebookresearch/theseus/compare/0.1.0-b.1...0.1.0-b.2
Source code(tar.gz)
Source code(zip)
0.1.0-b.1(Dec 3, 2021)
Initial beta release.

Core data structures.

Vector and 2D rigid body representations.

Gauss-Newton and LM nonlinear optimizers.

LU and Cholesky dense linear solvers.

Cholmod sparse linear solver (CPU only).

Cost functions for planar motion planning and tactile estimation in planar pushing.

Source code(tar.gz)
Source code(zip)