Repository: lucidrains/jax2torch
Branch: main
Commit: cd6c38a47827
Files: 7
Total size: 7.5 KB

Directory structure:
gitextract_43czjfz8/

├── .github/
│   └── workflows/
│       └── python-publish.yml
├── .gitignore
├── LICENSE
├── README.md
├── jax2torch/
│   ├── __init__.py
│   └── jax2torch.py
└── setup.py

================================================
FILE CONTENTS
================================================

================================================
FILE: .github/workflows/python-publish.yml
================================================
# 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 }}


================================================
FILE: .gitignore
================================================
# 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/
pip-wheel-metadata/
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/

# 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
target/

# Jupyter Notebook
.ipynb_checkpoints

# IPython
profile_default/
ipython_config.py

# pyenv
.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/


================================================
FILE: LICENSE
================================================
MIT License

Copyright (c) 2021 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.


================================================
FILE: README.md
================================================
## jax2torch

Use Jax functions in Pytorch with DLPack, as outlined <a href="https://gist.github.com/mattjj/e8b51074fed081d765d2f3ff90edf0e9">in a gist</a> by <a href="https://github.com/mattjj">@mattjj</a>. The repository was made for the purposes of making this <a href="https://github.com/spetti/SMURF">differentiable alignment work</a> interoperable with Pytorch projects.

## Install

```bash
$ pip install jax2torch
```

## Memory management

By default, Jax pre-allocates 90% of VRAM, which leaves Pytorch with very little left over.  To prevent this behavior, set the `XLA_PYTHON_CLIENT_PREALLOCATE` environmental variable to false before running any Jax code:

```python
import os
os.environ["XLA_PYTHON_CLIENT_PREALLOCATE"] = "false"
```

## Usage

[![Open In Colab](https://colab.research.google.com/assets/colab-badge.svg)](https://colab.research.google.com/drive/1GBEEnpuCvLS1bhb_xGCO5Y40rFiQrh6G?usp=sharing) Quick test

```python
import jax
import torch
from jax2torch import jax2torch
import os

os.environ["XLA_PYTHON_CLIENT_PREALLOCATE"] = "false"

# Jax function

@jax.jit
def jax_pow(x, y = 2):
  return x ** y

# convert to Torch function

torch_pow = jax2torch(jax_pow)

# run it on Torch data!

x = torch.tensor([1., 2., 3.])
y = torch_pow(x, y = 3)
print(y)  # tensor([1., 8., 27.])

# And differentiate!

x = torch.tensor([2., 3.], requires_grad = True)
y = torch.sum(torch_pow(x, y = 3))
y.backward()
print(x.grad) # tensor([12., 27.])
```


================================================
FILE: jax2torch/__init__.py
================================================
from jax2torch.jax2torch import jax2torch


================================================
FILE: jax2torch/jax2torch.py
================================================
# https://gist.github.com/mattjj/e8b51074fed081d765d2f3ff90edf0e9

import torch
from torch.utils import dlpack as torch_dlpack

import jax
from jax import dlpack as jax_dlpack
import jax.numpy as jnp
from jax.tree_util import tree_map

from inspect import signature
from functools import wraps

def j2t(x_jax):
    x_torch = torch_dlpack.from_dlpack(jax_dlpack.to_dlpack(x_jax))
    return x_torch

def t2j(x_torch):
    x_torch = x_torch.contiguous() # https://github.com/google/jax/issues/8082
    x_jax = jax_dlpack.from_dlpack(torch_dlpack.to_dlpack(x_torch))
    return x_jax

def tree_t2j(x_torch):
    return tree_map(lambda t: t2j(t) if isinstance(t, torch.Tensor) else t, x_torch)

def tree_j2t(x_jax):
    return tree_map(lambda t: j2t(t) if isinstance(t, jnp.ndarray) else t, x_jax)

def jax2torch(fn):
    @wraps(fn)
    def inner(*args, **kwargs):
        class JaxFun(torch.autograd.Function):
            @staticmethod
            def forward(ctx, *args):
                args = tree_t2j(args)
                y_, ctx.fun_vjp = jax.vjp(fn, *args)
                return tree_j2t(y_)

            @staticmethod
            def backward(ctx, *grad_args):
                grad_args = tree_t2j(grad_args) if len(grad_args) > 1 else t2j(grad_args[0])
                grads = ctx.fun_vjp(grad_args)
                grads = tuple(map(lambda t: t if isinstance(t, jnp.ndarray) else None, grads))
                return tree_j2t(grads)

        sig = signature(fn)
        bound = sig.bind(*args, **kwargs)
        bound.apply_defaults()
        return JaxFun.apply(*bound.arguments.values())
    return inner


================================================
FILE: setup.py
================================================
from setuptools import setup, find_packages

setup(
  name = 'jax2torch',
  packages = find_packages(exclude=[]),
  version = '0.0.7',
  license='MIT',
  description = 'Jax 2 Torch',
  author = 'Phil Wang',
  author_email = 'lucidrains@gmail.com',
  url = 'https://github.com/lucidrains/jax2torch',
  keywords = [
    'jax',
    'pytorch'
  ],
  install_requires=[
    'torch>=1.6',
    'jax>=0.2.20'
  ],
  classifiers=[
    'Development Status :: 4 - Beta',
    'Intended Audience :: Developers',
    'Topic :: Scientific/Engineering :: Artificial Intelligence',
    'License :: OSI Approved :: MIT License',
    'Programming Language :: Python :: 3.6',
  ],
)