rectified-flow-pytorch 0.0.1__tar.gz

rectified_flow_pytorch-0.0.1/.github/workflows/python-publish.yml

@@ -0,0 +1,36 @@
+# This workflow will upload a Python Package using Twine when a release is created
+# For more information see: https://help.github.com/en/actions/language-and-framework-guides/using-python-with-github-actions#publishing-to-package-registries
+
+# This workflow uses actions that are not certified by GitHub.
+# They are provided by a third-party and are governed by
+# separate terms of service, privacy policy, and support
+# documentation.
+
+name: Upload Python Package
+
+on:
+  release:
+    types: [published]
+
+jobs:
+  deploy:
+
+    runs-on: ubuntu-latest
+
+    steps:
+    - uses: actions/checkout@v2
+    - name: Set up Python
+      uses: actions/setup-python@v2
+      with:
+        python-version: '3.x'
+    - name: Install dependencies
+      run: |
+        python -m pip install --upgrade pip
+        pip install build
+    - name: Build package
+      run: python -m build
+    - name: Publish package
+      uses: pypa/gh-action-pypi-publish@27b31702a0e7fc50959f5ad993c78deac1bdfc29
+      with:
+        user: __token__
+        password: ${{ secrets.PYPI_API_TOKEN }}

rectified_flow_pytorch-0.0.1/.gitignore

@@ -0,0 +1,162 @@
+# 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
+
+# poetry
+#   Similar to Pipfile.lock, it is generally recommended to include poetry.lock in version control.
+#   This is especially recommended for binary packages to ensure reproducibility, and is more
+#   commonly ignored for libraries.
+#   https://python-poetry.org/docs/basic-usage/#commit-your-poetrylock-file-to-version-control
+#poetry.lock
+
+# pdm
+#   Similar to Pipfile.lock, it is generally recommended to include pdm.lock in version control.
+#pdm.lock
+#   pdm stores project-wide configurations in .pdm.toml, but it is recommended to not include it
+#   in version control.
+#   https://pdm.fming.dev/latest/usage/project/#working-with-version-control
+.pdm.toml
+.pdm-python
+.pdm-build/
+
+# PEP 582; used by e.g. github.com/David-OConnor/pyflow and github.com/pdm-project/pdm
+__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/
+
+# PyCharm
+#  JetBrains specific template is maintained in a separate JetBrains.gitignore that can
+#  be found at https://github.com/github/gitignore/blob/main/Global/JetBrains.gitignore
+#  and can be added to the global gitignore or merged into this file.  For a more nuclear
+#  option (not recommended) you can uncomment the following to ignore the entire idea folder.
+#.idea/

rectified_flow_pytorch-0.0.1/LICENSE

@@ -0,0 +1,21 @@
+MIT License
+
+Copyright (c) 2024 Phil Wang
+
+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.

rectified_flow_pytorch-0.0.1/PKG-INFO

@@ -0,0 +1,138 @@
+Metadata-Version: 2.3
+Name: rectified-flow-pytorch
+Version: 0.0.1
+Summary: Rectified Flow in Pytorch
+Project-URL: Homepage, https://pypi.org/project/rectified-flow-pytorch/
+Project-URL: Repository, https://github.com/lucidrains/rectified-flow-pytorch
+Author-email: Phil Wang <lucidrains@gmail.com>
+License: MIT License
+        
+        Copyright (c) 2024 Phil Wang
+        
+        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.
+License-File: LICENSE
+Keywords: artificial intelligence,deep learning,rectified flow
+Classifier: Development Status :: 4 - Beta
+Classifier: Intended Audience :: Developers
+Classifier: License :: OSI Approved :: MIT License
+Classifier: Programming Language :: Python :: 3.8
+Classifier: Topic :: Scientific/Engineering :: Artificial Intelligence
+Requires-Python: >=3.8
+Requires-Dist: einops>=0.8.0
+Requires-Dist: scipy
+Requires-Dist: torch>=2.0
+Requires-Dist: torchdiffeq
+Provides-Extra: examples
+Description-Content-Type: text/markdown
+
+## Rectified Flow - Pytorch (wip)
+
+Implementation of rectified flow and some of its followup research / improvements in Pytorch
+
+## Install
+
+```bash
+$ pip install rectified-flow-pytorch
+```
+
+## Usage
+
+```python
+import torch
+from torch import nn
+
+from rectified_flow_pytorch import RectifiedFlow
+
+model = nn.Conv2d(3, 3, 1)
+
+rectified_flow = RectifiedFlow(model, time_cond_kwarg = None)
+
+images = torch.randn(1, 3, 256, 256)
+
+loss = rectified_flow(images)
+loss.backward()
+
+sampled = rectified_flow.sample()
+assert sampled.shape == images.shape
+```
+
+For reflow as described in the paper
+
+```python
+import torch
+from torch import nn
+
+from rectified_flow_pytorch import RectifiedFlow, Reflow
+
+model = nn.Conv2d(3, 3, 1)
+
+rectified_flow = RectifiedFlow(model, time_cond_kwarg = None)
+
+images = torch.randn(1, 3, 256, 256)
+
+loss = rectified_flow(images)
+loss.backward()
+
+# first train on many images
+
+reflow = Reflow(rectified_flow)
+
+reflow_loss = reflow()
+reflow_loss.backward()
+
+# then do the above in a loop many times
+
+sampled = reflow.sample()
+assert sampled.shape == images.shape
+```
+
+## Citations
+
+```bibtex
+@article{Liu2022FlowSA,
+    title   = {Flow Straight and Fast: Learning to Generate and Transfer Data with Rectified Flow},
+    author  = {Xingchao Liu and Chengyue Gong and Qiang Liu},
+    journal = {ArXiv},
+    year    = {2022},
+    volume  = {abs/2209.03003},
+    url     = {https://api.semanticscholar.org/CorpusID:252111177}
+}
+```
+
+```bibtex
+@article{Lee2024ImprovingTT,
+    title   = {Improving the Training of Rectified Flows},
+    author  = {Sangyun Lee and Zinan Lin and Giulia Fanti},
+    journal = {ArXiv},
+    year    = {2024},
+    volume  = {abs/2405.20320},
+    url     = {https://api.semanticscholar.org/CorpusID:270123378}
+}
+```
+
+```bibtex
+@article{Esser2024ScalingRF,
+    title   = {Scaling Rectified Flow Transformers for High-Resolution Image Synthesis},
+    author  = {Patrick Esser and Sumith Kulal and A. Blattmann and Rahim Entezari and Jonas Muller and Harry Saini and Yam Levi and Dominik Lorenz and Axel Sauer and Frederic Boesel and Dustin Podell and Tim Dockhorn and Zion English and Kyle Lacey and Alex Goodwin and Yannik Marek and Robin Rombach},
+    journal = {ArXiv},
+    year    = {2024},
+    volume  = {abs/2403.03206},
+    url     = {https://api.semanticscholar.org/CorpusID:268247980}
+}
+```

rectified_flow_pytorch-0.0.1/README.md

@@ -0,0 +1,95 @@
+## Rectified Flow - Pytorch (wip)
+
+Implementation of rectified flow and some of its followup research / improvements in Pytorch
+
+## Install
+
+```bash
+$ pip install rectified-flow-pytorch
+```
+
+## Usage
+
+```python
+import torch
+from torch import nn
+
+from rectified_flow_pytorch import RectifiedFlow
+
+model = nn.Conv2d(3, 3, 1)
+
+rectified_flow = RectifiedFlow(model, time_cond_kwarg = None)
+
+images = torch.randn(1, 3, 256, 256)
+
+loss = rectified_flow(images)
+loss.backward()
+
+sampled = rectified_flow.sample()
+assert sampled.shape == images.shape
+```
+
+For reflow as described in the paper
+
+```python
+import torch
+from torch import nn
+
+from rectified_flow_pytorch import RectifiedFlow, Reflow
+
+model = nn.Conv2d(3, 3, 1)
+
+rectified_flow = RectifiedFlow(model, time_cond_kwarg = None)
+
+images = torch.randn(1, 3, 256, 256)
+
+loss = rectified_flow(images)
+loss.backward()
+
+# first train on many images
+
+reflow = Reflow(rectified_flow)
+
+reflow_loss = reflow()
+reflow_loss.backward()
+
+# then do the above in a loop many times
+
+sampled = reflow.sample()
+assert sampled.shape == images.shape
+```
+
+## Citations
+
+```bibtex
+@article{Liu2022FlowSA,
+    title   = {Flow Straight and Fast: Learning to Generate and Transfer Data with Rectified Flow},
+    author  = {Xingchao Liu and Chengyue Gong and Qiang Liu},
+    journal = {ArXiv},
+    year    = {2022},
+    volume  = {abs/2209.03003},
+    url     = {https://api.semanticscholar.org/CorpusID:252111177}
+}
+```
+
+```bibtex
+@article{Lee2024ImprovingTT,
+    title   = {Improving the Training of Rectified Flows},
+    author  = {Sangyun Lee and Zinan Lin and Giulia Fanti},
+    journal = {ArXiv},
+    year    = {2024},
+    volume  = {abs/2405.20320},
+    url     = {https://api.semanticscholar.org/CorpusID:270123378}
+}
+```
+
+```bibtex
+@article{Esser2024ScalingRF,
+    title   = {Scaling Rectified Flow Transformers for High-Resolution Image Synthesis},
+    author  = {Patrick Esser and Sumith Kulal and A. Blattmann and Rahim Entezari and Jonas Muller and Harry Saini and Yam Levi and Dominik Lorenz and Axel Sauer and Frederic Boesel and Dustin Podell and Tim Dockhorn and Zion English and Kyle Lacey and Alex Goodwin and Yannik Marek and Robin Rombach},
+    journal = {ArXiv},
+    year    = {2024},
+    volume  = {abs/2403.03206},
+    url     = {https://api.semanticscholar.org/CorpusID:268247980}
+}
+```

rectified_flow_pytorch-0.0.1/pyproject.toml

@@ -0,0 +1,46 @@
+[project]
+name = "rectified-flow-pytorch"
+version = "0.0.1"
+description = "Rectified Flow in Pytorch"
+authors = [
+    { name = "Phil Wang", email = "lucidrains@gmail.com" }
+]
+readme = "README.md"
+requires-python = ">= 3.8"
+license = { file = "LICENSE" }
+keywords = [
+    'artificial intelligence',
+    'deep learning',
+    'rectified flow'
+  ]
+classifiers=[
+    'Development Status :: 4 - Beta',
+    'Intended Audience :: Developers',
+    'Topic :: Scientific/Engineering :: Artificial Intelligence',
+    'License :: OSI Approved :: MIT License',
+    'Programming Language :: Python :: 3.8',
+  ]
+
+dependencies = [
+    'einops>=0.8.0',
+    'scipy',
+    'torch>=2.0',
+    'torchdiffeq',
+]
+
+[project.urls]
+Homepage = "https://pypi.org/project/rectified-flow-pytorch/"
+Repository = "https://github.com/lucidrains/rectified-flow-pytorch"
+
+[project.optional-dependencies]
+examples = []
+
+[build-system]
+requires = ["hatchling"]
+build-backend = "hatchling.build"
+
+[tool.hatch.metadata]
+allow-direct-references = true
+
+[tool.hatch.build.targets.wheel]
+packages = ["rectified_flow_pytorch"]

rectified_flow_pytorch-0.0.1/rectified_flow_pytorch/__init__.py

@@ -0,0 +1 @@
+from rectified_flow_pytorch.rectified_flow import RectifiedFlow, Reflow

rectified_flow_pytorch-0.0.1/rectified_flow_pytorch/rectified_flow.py

@@ -0,0 +1,168 @@
+from __future__ import annotations
+from typing import Tuple
+from copy import deepcopy
+
+import torch
+from torch.nn import Module
+import torch.nn.functional as F
+
+from torchdiffeq import odeint
+
+# helpers
+
+def exists(v):
+    return v is not None
+
+def default(v, d):
+    return v if exists(v) else d
+
+# tensor helpers
+
+def append_dims(t, ndims):
+    shape = t.shape
+    return t.reshape(*shape, *((1,) * ndims))
+
+# main class
+
+class RectifiedFlow(Module):
+    def __init__(
+        self,
+        model: Module,
+        time_cond_kwarg: str | None = 'times',
+        odeint_kwargs: dict = dict(
+            atol = 1e-5,
+            rtol = 1e-5,
+            method = 'midpoint'
+        ),
+        data_shape: Tuple[int, ...] | None = None,
+    ):
+        super().__init__()
+        self.model = model
+        self.time_cond_kwarg = time_cond_kwarg # whether the model is to be conditioned on the times
+
+        # sampling
+
+        self.odeint_kwargs = odeint_kwargs
+        self.data_shape = data_shape
+
+    @property
+    def device(self):
+        return next(self.model.parameters()).device
+
+    @torch.no_grad()
+    def sample(
+        self,
+        batch_size = 1,
+        steps = 16,
+        noise = None,
+        data_shape: Tuple[int, ...] | None = None,
+        **model_kwargs
+    ):
+        was_training = self.training
+        self.eval()
+
+        data_shape = default(data_shape, self.data_shape)
+        assert exists(data_shape), 'you need to either pass in a `data_shape` or have trained at least with one forward'
+
+        def ode_fn(t, x):
+            time_kwarg = self.time_cond_kwarg
+
+            if exists(time_kwarg):
+                model_kwargs.update(**{time_kwarg: t})
+
+            return self.model(x, **model_kwargs)
+
+        # start with random gaussian noise - y0
+
+        noise = default(noise, torch.randn((batch_size, *data_shape)))
+
+        # time steps
+
+        times = torch.linspace(0., 1., steps, device = self.device)
+
+        # ode
+
+        trajectory = odeint(ode_fn, noise, times, **self.odeint_kwargs)
+
+        sampled_data = trajectory[-1]
+
+        self.train(was_training)
+        return sampled_data
+
+    def forward(
+        self,
+        data,
+        noise = None,
+        **model_kwargs
+    ):
+        batch, *data_shape = data.shape
+
+        self.data_shape = default(self.data_shape, data_shape)
+
+        # x0 - gaussian noise, x1 - data
+
+        noise = default(noise, torch.randn_like(data))
+
+        # times, and times with dimension padding on right
+
+        times = torch.rand(batch, device = self.device)
+        padded_times = append_dims(times, data.ndim - 1)
+
+        # Algorithm 2 in paper
+        # linear interpolation of noise with data using random times
+        # x1 * t + x0 * (1 - t) - so from noise (time = 0) to data (time = 1.)
+
+        noised = padded_times * data + (1. - padded_times) * noise
+
+        # prepare maybe time conditioning for model
+        
+        time_kwarg = self.time_cond_kwarg
+
+        if exists(time_kwarg):
+            model_kwargs.update(**{time_kwarg: times})
+
+        # the model predicts the flow from the noised data
+
+        flow = data - noise
+        pred_flow = self.model(noised, **model_kwargs)
+
+        loss = F.mse_loss(pred_flow, flow)
+
+        return loss
+
+# reflow wrapper
+
+class Reflow(Module):
+    def __init__(
+        self,
+        rectified_flow: RectifiedFlow,
+        *,
+        batch_size = 16,
+
+    ):
+        super().__init__()
+        model, data_shape = rectified_flow.model, rectified_flow.data_shape
+        assert exists(data_shape), '`data_shape` must be defined in RectifiedFlow'
+
+        self.batch_size = batch_size
+        self.data_shape = data_shape
+
+        self.model = rectified_flow
+        self.frozen_model = deepcopy(rectified_flow)
+
+    def parameters(self):
+        return self.model.parameters() # omit frozen model
+
+    def sample(self, *args, **kwargs):
+        return self.model.sample(*args, **kwargs)
+
+    def forward(self):
+
+        noise = torch.randn((self.batch_size, *self.data_shape))
+        sampled_output = self.frozen_model.sample(noise = noise)
+
+        # the coupling in the paper is (noise, sampled_output)
+
+        loss = self.model(sampled_output, noise = noise)
+
+        return loss