init

.gitignore

@@ -0,0 +1,146 @@
+!.gitkeep
+
+# Created by https://www.toptal.com/developers/gitignore/api/python
+# Edit at https://www.toptal.com/developers/gitignore?templates=python
+
+### Python ###
+# Byte-compiled / optimized / DLL files
+__pycache__/
+*.py[cod]
+*$py.class
+
+# C extensions
+*.so
+
+# Distribution / packaging
+.Python
+build/
+develop-eggs/
+dist/
+downloads/
+eggs/
+.eggs/
+lib/
+lib64/
+parts/
+sdist/
+var/
+wheels/
+share/python-wheels/
+*.egg-info/
+.installed.cfg
+*.egg
+MANIFEST
+
+# PyInstaller
+#  Usually these files are written by a python script from a template
+#  before PyInstaller builds the exe, so as to inject date/other infos into it.
+*.manifest
+*.spec
+
+# Installer logs
+pip-log.txt
+pip-delete-this-directory.txt
+
+# Unit test / coverage reports
+htmlcov/
+.tox/
+.nox/
+.coverage
+.coverage.*
+.cache
+nosetests.xml
+coverage.xml
+*.cover
+*.py,cover
+.hypothesis/
+.pytest_cache/
+cover/
+
+# Translations
+*.mo
+*.pot
+
+# Django stuff:
+*.log
+local_settings.py
+db.sqlite3
+db.sqlite3-journal
+
+# Flask stuff:
+instance/
+.webassets-cache
+
+# Scrapy stuff:
+.scrapy
+
+# Sphinx documentation
+docs/_build/
+
+# PyBuilder
+.pybuilder/
+target/
+
+# Jupyter Notebook
+.ipynb_checkpoints
+
+# IPython
+profile_default/
+ipython_config.py
+
+# pyenv
+#   For a library or package, you might want to ignore these files since the code is
+#   intended to run in multiple environments; otherwise, check them in:
+# .python-version
+
+# pipenv
+#   According to pypa/pipenv#598, it is recommended to include Pipfile.lock in version control.
+#   However, in case of collaboration, if having platform-specific dependencies or dependencies
+#   having no cross-platform support, pipenv may install dependencies that don't work, or not
+#   install all needed dependencies.
+#Pipfile.lock
+
+# PEP 582; used by e.g. github.com/David-OConnor/pyflow
+__pypackages__/
+
+# Celery stuff
+celerybeat-schedule
+celerybeat.pid
+
+# SageMath parsed files
+*.sage.py
+
+# Environments
+.env
+.venv
+env/
+venv/
+ENV/
+env.bak/
+venv.bak/
+
+# Spyder project settings
+.spyderproject
+.spyproject
+
+# Rope project settings
+.ropeproject
+
+# mkdocs documentation
+/site
+
+# mypy
+.mypy_cache/
+.dmypy.json
+dmypy.json
+
+# Pyre type checker
+.pyre/
+
+# pytype static type analyzer
+.pytype/
+
+# Cython debug symbols
+cython_debug/
+
+# End of https://www.toptal.com/developers/gitignore/api/python

LICENSE

@@ -0,0 +1,21 @@
+MIT License
+
+Copyright (c) 2021 Mark Tuddenham
+
+Permission is hereby granted, free of charge, to any person obtaining a copy
+of this software and associated documentation files (the "Software"), to deal
+in the Software without restriction, including without limitation the rights
+to use, copy, modify, merge, publish, distribute, sublicense, and/or sell
+copies of the Software, and to permit persons to whom the Software is
+furnished to do so, subject to the following conditions:
+
+The above copyright notice and this permission notice shall be included in all
+copies or substantial portions of the Software.
+
+THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, EXPRESS OR
+IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF MERCHANTABILITY,
+FITNESS FOR A PARTICULAR PURPOSE AND NONINFRINGEMENT. IN NO EVENT SHALL THE
+AUTHORS OR COPYRIGHT HOLDERS BE LIABLE FOR ANY CLAIM, DAMAGES OR OTHER
+LIABILITY, WHETHER IN AN ACTION OF CONTRACT, TORT OR OTHERWISE, ARISING FROM,
+OUT OF OR IN CONNECTION WITH THE SOFTWARE OR THE USE OR OTHER DEALINGS IN THE
+SOFTWARE.

README.md

@@ -0,0 +1,39 @@
+# Orthogonalised Optimisers
+
+## Install  package
+```bash
+pip install .
+```
+or 
+
+```bash
+pip install git+https://github.com/MarkTuddenham/Orthogonal-Optimiser.git#egg=orth_optim
+```
+
+## Usage
+And then at the top of your main python script:
+
+```python
+from orth_optim import hook
+hook()
+```
+Now SGD has an orthogonal option
+```python
+torch.optim.SGD(model.parameters(),
+                lr=1e-3,
+                momentum=0.9,
+                orth=True)
+```
+
+## Custom Optimisers
+If you have a custom optimiser you can apply the `orthogonalise` decorator.
+
+```python
+from orth_optim import orthogonalise
+
+@orthogonalise
+class LARS(torch.optim.Optimizer):
+	...
+```
+
+

setup.py

@@ -0,0 +1,33 @@
+#!/usr/bin/env python
+
+import os
+from setuptools import setup
+from setuptools import find_packages
+
+# User-friendly description from README.md
+current_directory = os.path.dirname(os.path.abspath(__file__))
+try:
+    with open(os.path.join(current_directory, 'README.md'), encoding='utf-8') as f:
+        long_description = f.read()
+except Exception:
+    long_description = ''
+
+setup(name='orth_optim',
+      author='Mark Tuddenham',
+      author_email='[email protected]',
+      version='1.0.0',
+      description='Add option to orthogonalise gradients to pytorch optimisers',
+      long_description=long_description,
+      long_description_context_type='text/markdown',
+      license='MIT',
+      url='https://github.com/MarkTuddenham/Orthogonal-Optimisers',
+      keywords=['optimisation', 'optimzation', 'pytorch'],
+      packages=find_packages(where='src'),
+      package_dir={"": "src"},
+      install_requires=['torch'],
+      classifiers=[
+          "Programming Language :: Python :: 3",
+          "License :: OSI Approved :: MIT License",
+          "Operating System :: OS Independent",
+      ]
+      )

src/orth_optim/__init__.py

@@ -0,0 +1,74 @@
+from typing import TypeVar
+from typing import Type
+from typing import Any
+
+import logging
+
+import torch
+
+OptimType = TypeVar('U', bound=torch.optim.Optimizer)
+
+logger: logging.Logger = logging.getLogger(__name__)
+logger.setLevel(logging.DEBUG)
+logger.addHandler(logging.NullHandler())
+
+
+@torch.no_grad()
+def _orth_grads(optimiser: OptimType) -> None:
+    # Orthogonalise the gradients using SVD
+    for group in optimiser.param_groups:
+        orth = group['orth']
+        for i, p in enumerate(group['params']):
+            if orth and p.grad is not None and p.ndim > 1:
+                G: torch.Tensor = p.grad.flatten(start_dim=1)
+                try:
+                    u, s, vt = torch.linalg.svd(G, full_matrices=False)
+                    orth_G: torch.Tensor = u @ vt
+                except RuntimeError:
+                    logger.error('Failed to perform SVD, adding some noise.')
+                    try:
+                        u, s, v = torch.svd_lowrank(
+                            G,
+                            q=1,    # assume rank is at least 1
+                            M=1e-4 * G.mean() * torch.randn_like(G))
+                        orth_G: torch.Tensor = u @ v.T
+                    except RuntimeError:
+                        logger.error(('Failed to perform SVD with noise,'
+                                      ' skipping gradient orthogonalisation'))
+                        return
+                p.grad = orth_G.reshape_as(p)
+
+
+def orthogonalise(cls: Type[OptimType]) -> Type[OptimType]:
+    og_init = cls.__init__
+    og_step = cls.step
+
+    def new_init(self, *args, orth=False, **kwargs):
+        # Add orth hyperparam to defaults
+        og_init(self, *args, **kwargs)
+        self.defaults['orth'] = orth
+        for grp in self.param_groups:
+            grp.setdefault('orth', orth)
+
+    def new_step(self, *args, **kwargs):
+        # Orthogonalise the grads before the original optim's step method
+        _orth_grads(self)
+        og_step(self, *args, **kwargs)
+
+    cls.__init__ = new_init
+    cls.step = new_step
+    return cls
+
+
+def hook() -> None:
+    from inspect import isclass
+    for mod in dir(torch.optim):
+        if mod.startswith('_'):
+            continue
+        _optim: Any = getattr(torch.optim, mod)
+        if (
+            isclass(_optim)
+            and issubclass(_optim, torch.optim.Optimizer)
+            and _optim is not torch.optim.Optimizer
+        ):
+            setattr(torch.optim, mod, orthogonalise(_optim))