# torchtree-physher
[](https://github.com/4ment/torchtree-physher/actions/workflows/test_linux.yml)
[](https://www.gnu.org/licenses/gpl-3.0)
## About torchtree-physher
torchtree-physher is a python package providing fast gradient calculation implemented in [physher] for [torchtree].
## Getting Started
A C++ compiler such as ``g++`` or ``clang++`` is required.
On Debian-based systems, this can be installed via ``apt``:
```bash
sudo apt install g++
```
On MacOS, it is recommended to use the latest version of ``clang++``:
```bash
brew install llvm
```
The [pybind11] library is also used for binding the
C++ functionality to Python.
### Dependencies
- [physher]
- [pybind11]
- [PyTorch]
- [torchtree]
[physher] is a phylogenetic program written in C that provides C++ wrappers to compute the tree and coalescent likelihoods and their gradients under different models.
To build physher from source you can run
```bash
git clone https://github.com/4ment/physher
cmake -S physher/ -B physher/build -DBUILD_CPP_WRAPPER=on -DBUILD_TESTING=on
cmake --build physher/build/ --target install
```
Check it works (optional)
```bash
ctest --test-dir physher/build/
```
### Installation
To build `torchtree-physher` from source you can run
```bash
git clone https://github.com/4ment/torchtree-physher
pip install torchtree-physher/
```
### Check install
If the installation was successful, this command should print the version of the `torchtree_physher` library
```bash
python -c "import torchtree_physher;print(torchtree_physher.__version__)"
```
## Command line arguments
The torchtree-physher plugin adds these arguments to the torchtree CLI:
```bash
torchtree-cli advi --help
...
--physher use physher
--physher_include_jacobian
include Jacobian of the node height transform in the node height gradient
--physher_disable_sse
disable SSE in physher
--physher_disable_coalescent
disable coalescent calculation by physher
--physher_site {weibull,gamma}
distribution for rate heterogeneity across sites
```
## Features
### Tree likelihood
Some types in the JSON configuration file have to be replaced in order to use the tree likelihood implementation of physher. You simply need to add `torchtree_physher.` before a model type. Here is a list of models implemented in this plugin:
- `TreeLikelihoodModel`
- Substitution models:
- `JC69`
- `HKY`
- `GTR`
- `GeneralNonSymmetricSubstitutionModel`
- Tree models:
- `UnRootedTreeModel`
- `ReparameterizedTimeTreeModel`
- Clock models (optional):
- `StrictClockModel`
- `SimpleClockModel`
- Site models:
- `ConstantSiteModel`
- `GammaSiteModel`
- `InvariantSiteModel`
- `WeibullSiteModel`
Note that the type of every sub-model of the tree likelihood object (e.g. site, tree models...) has to be replaced.
For example if we want to use ADVI with an unrooted tree and a Weibull site model:
```bash
torchtree-cli advi -i data.fa -t data.tree -C 4 > data.json
sed -i -E 's/TreeLikelihoodModel/torchtree_physher.TreeLikelihoodModel/; s/UnRootedTreeModel/torchtree_physher.UnRootedTreeModel/; s/WeibullSiteModel/torchtree_physher.WeibullSiteModel/' data.json
torchtree data.json
```
The JSON file can be created directly using the `--physher` option:
```bash
torchtree-cli advi -i data.fa -t data.tree -C 4 --physher > data.json
```
### Coalescent models
Here is a list of coalescent models implemented in this plugin:
- `ConstantCoalescentModel`
- `PiecewiseConstantCoalescentGridModel` (aka skygrid)
- `PiecewiseConstantCoalescentModel` (aka skyride)
- `PiecewiseLinearCoalescentGridModel`
## License
Distributed under the GPLv3 License. See [LICENSE](LICENSE) for more information.
## Acknowledgements
torchtree-physher makes use of the following libraries and tools, which are under their own respective licenses:
- [physher]
- [pybind11]
- [PyTorch]
- [torchtree]
[physher]: https://github.com/4ment/physher
[pybind11]: https://pybind11.readthedocs.io/en/stable
[PyTorch]: https://pytorch.org
[torchtree]: https://github.com/4ment/torchtree
Raw data
{
"_id": null,
"home_page": "https://github.com/4ment/torchtree-physher",
"name": "torchtree-physher",
"maintainer": null,
"docs_url": null,
"requires_python": ">=3.7",
"maintainer_email": null,
"keywords": "phylogenetics, variational, Bayes, pytorch",
"author": "Mathieu Fourment",
"author_email": "mathieu.fourment@uts.edu.au",
"download_url": "https://files.pythonhosted.org/packages/4a/f2/aeef673bca12a5bc12a5d97895f9406e919c4dc39c63e2ac9a11ca8aee3a/torchtree-physher-1.0.0.tar.gz",
"platform": null,
"description": "# torchtree-physher\n\n[](https://github.com/4ment/torchtree-physher/actions/workflows/test_linux.yml)\n[](https://www.gnu.org/licenses/gpl-3.0)\n\n## About torchtree-physher\ntorchtree-physher is a python package providing fast gradient calculation implemented in [physher] for [torchtree].\n\n## Getting Started\n\nA C++ compiler such as ``g++`` or ``clang++`` is required.\nOn Debian-based systems, this can be installed via ``apt``:\n\n```bash\nsudo apt install g++\n```\n\nOn MacOS, it is recommended to use the latest version of ``clang++``:\n```bash\nbrew install llvm\n```\n\nThe [pybind11] library is also used for binding the\nC++ functionality to Python.\n\n### Dependencies\n - [physher]\n - [pybind11]\n - [PyTorch]\n - [torchtree]\n\n[physher] is a phylogenetic program written in C that provides C++ wrappers to compute the tree and coalescent likelihoods and their gradients under different models.\n\nTo build physher from source you can run\n```bash\ngit clone https://github.com/4ment/physher\ncmake -S physher/ -B physher/build -DBUILD_CPP_WRAPPER=on -DBUILD_TESTING=on\ncmake --build physher/build/ --target install\n```\n\nCheck it works (optional)\n```bash\nctest --test-dir physher/build/\n```\n\n\n### Installation\nTo build `torchtree-physher` from source you can run\n```bash\ngit clone https://github.com/4ment/torchtree-physher\npip install torchtree-physher/\n```\n\n### Check install\nIf the installation was successful, this command should print the version of the `torchtree_physher` library\n```bash\npython -c \"import torchtree_physher;print(torchtree_physher.__version__)\"\n```\n\n## Command line arguments\nThe torchtree-physher plugin adds these arguments to the torchtree CLI:\n\n```bash\ntorchtree-cli advi --help\n ...\n --physher use physher\n --physher_include_jacobian\n include Jacobian of the node height transform in the node height gradient\n --physher_disable_sse\n disable SSE in physher\n --physher_disable_coalescent\n disable coalescent calculation by physher\n --physher_site {weibull,gamma}\n distribution for rate heterogeneity across sites\n```\n\n## Features\n### Tree likelihood\nSome types in the JSON configuration file have to be replaced in order to use the tree likelihood implementation of physher. You simply need to add `torchtree_physher.` before a model type. Here is a list of models implemented in this plugin:\n\n- `TreeLikelihoodModel`\n- Substitution models:\n - `JC69`\n - `HKY`\n - `GTR`\n - `GeneralNonSymmetricSubstitutionModel`\n- Tree models:\n - `UnRootedTreeModel`\n - `ReparameterizedTimeTreeModel`\n- Clock models (optional):\n - `StrictClockModel`\n - `SimpleClockModel`\n- Site models:\n - `ConstantSiteModel`\n - `GammaSiteModel`\n - `InvariantSiteModel`\n - `WeibullSiteModel`\n\nNote that the type of every sub-model of the tree likelihood object (e.g. site, tree models...) has to be replaced.\n\nFor example if we want to use ADVI with an unrooted tree and a Weibull site model:\n\n```bash\ntorchtree-cli advi -i data.fa -t data.tree -C 4 > data.json\nsed -i -E 's/TreeLikelihoodModel/torchtree_physher.TreeLikelihoodModel/; s/UnRootedTreeModel/torchtree_physher.UnRootedTreeModel/; s/WeibullSiteModel/torchtree_physher.WeibullSiteModel/' data.json\ntorchtree data.json\n```\n\nThe JSON file can be created directly using the `--physher` option:\n```bash\ntorchtree-cli advi -i data.fa -t data.tree -C 4 --physher > data.json\n```\n\n### Coalescent models\nHere is a list of coalescent models implemented in this plugin:\n\n- `ConstantCoalescentModel`\n- `PiecewiseConstantCoalescentGridModel` (aka skygrid)\n- `PiecewiseConstantCoalescentModel` (aka skyride)\n- `PiecewiseLinearCoalescentGridModel`\n\n## License\n\nDistributed under the GPLv3 License. See [LICENSE](LICENSE) for more information.\n\n## Acknowledgements\n\ntorchtree-physher makes use of the following libraries and tools, which are under their own respective licenses:\n\n - [physher]\n - [pybind11]\n - [PyTorch]\n - [torchtree]\n\n[physher]: https://github.com/4ment/physher\n[pybind11]: https://pybind11.readthedocs.io/en/stable\n[PyTorch]: https://pytorch.org\n[torchtree]: https://github.com/4ment/torchtree\n",
"bugtrack_url": null,
"license": "GPL3",
"summary": "Phylogenetic inference with pytorch and physher",
"version": "1.0.0",
"project_urls": {
"Homepage": "https://github.com/4ment/torchtree-physher"
},
"split_keywords": [
"phylogenetics",
" variational",
" bayes",
" pytorch"
],
"urls": [
{
"comment_text": "",
"digests": {
"blake2b_256": "ef418c3be80ab9227f7a871847190340d4223b1643fa2cf7af42c2bfd47c273d",
"md5": "049951e05063a9950c5f32791b96f393",
"sha256": "ff497093f473c000c1e24ae25c76bda0f14161e8713a6117ff88e15669d46fda"
},
"downloads": -1,
"filename": "torchtree_physher-1.0.0-cp312-cp312-macosx_11_0_arm64.whl",
"has_sig": false,
"md5_digest": "049951e05063a9950c5f32791b96f393",
"packagetype": "bdist_wheel",
"python_version": "cp312",
"requires_python": ">=3.7",
"size": 176871,
"upload_time": "2024-07-10T05:13:15",
"upload_time_iso_8601": "2024-07-10T05:13:15.535918Z",
"url": "https://files.pythonhosted.org/packages/ef/41/8c3be80ab9227f7a871847190340d4223b1643fa2cf7af42c2bfd47c273d/torchtree_physher-1.0.0-cp312-cp312-macosx_11_0_arm64.whl",
"yanked": false,
"yanked_reason": null
},
{
"comment_text": "",
"digests": {
"blake2b_256": "4af2aeef673bca12a5bc12a5d97895f9406e919c4dc39c63e2ac9a11ca8aee3a",
"md5": "19931ccb574188d6b2089d88c4f7cfe7",
"sha256": "1f84064326b6e490871f0279243af62d78390bd6b67e2444d7825279f27323e9"
},
"downloads": -1,
"filename": "torchtree-physher-1.0.0.tar.gz",
"has_sig": false,
"md5_digest": "19931ccb574188d6b2089d88c4f7cfe7",
"packagetype": "sdist",
"python_version": "source",
"requires_python": ">=3.7",
"size": 32420,
"upload_time": "2024-07-10T05:13:17",
"upload_time_iso_8601": "2024-07-10T05:13:17.748037Z",
"url": "https://files.pythonhosted.org/packages/4a/f2/aeef673bca12a5bc12a5d97895f9406e919c4dc39c63e2ac9a11ca8aee3a/torchtree-physher-1.0.0.tar.gz",
"yanked": false,
"yanked_reason": null
}
],
"upload_time": "2024-07-10 05:13:17",
"github": true,
"gitlab": false,
"bitbucket": false,
"codeberg": false,
"github_user": "4ment",
"github_project": "torchtree-physher",
"travis_ci": false,
"coveralls": false,
"github_actions": true,
"lcname": "torchtree-physher"
}
JSON
