# cfdonnx
A Python module for exporting pre-trained CFD models to ONNX, making them interoperable with other ML frameworks and compatible with browsers.
It currently supports U-Net architecture and PyTorch models, but it will be soon extended to other frameworks and architectures.
Reproducible examples can be found at [openfoam-cfd-rom](https://github.com/simzero/openfoam-ml-rom) usign [DeepCFD](https://github.com/mdribeiro/DeepCFD).
## Installation
The module can be installed with:
```
pip3 install cfdonnx
```
## Usage
```
Usage: python3 -m cfdonnx [OPTIONS]
Options:
-n, --net TEXT network architecture: UNetEx or AutoEncoder (default: UNetEx)
-i, --input PATH checkpoint (default: checkpoint.pt)
-o, --output PATH ONNX output file (default: checkpoint.onnx)
-k, --kernel-size INT kernel size (optional, read from state_dict['kernel_size] by default )
-f, --filters TEXT filter size, e.g. 8,16,32,32 (optional, read from state_dict['filters'] by default)
-c --channels INT number of channels (optional, read from state_dict['input_shape'] by default)
-x --nx INT X dimension (optional, read from state_dict['input_shape'] by default)
-y --ny INT Y dimension (optional, read from state_dict['input_shape'] by default )
-o, --output PATH Save model path (default: mymodel.pt)
Example:
python3 -m cfdonnx \
--net UNetEx \
--input flowAroundObstacles.pt \
--output flowAroundObstacles.onnx
```
You can use your CFD ONNX models on runtime in Babylon.js as showcased at https://play.simzero.com/#D3SFTH#6 for the [flowAroundObstacles](https://github.com/simzero/openfoam-ml-rom/tree/main/OpenFOAM/incompressible/simpleFoam/flowAroundObstacles) example.
A generic template for using ONNX is also available at https://play.simzero.com/#WIB297#1.
Raw data
{
"_id": null,
"home_page": "https://github.com/simzero/cfdonnx",
"name": "cfdonnx",
"maintainer": "",
"docs_url": null,
"requires_python": "",
"maintainer_email": "",
"keywords": "cfd ml onnx",
"author": "Carlos Pena Monferrer",
"author_email": "",
"download_url": "https://files.pythonhosted.org/packages/db/63/da3c5c60e697c1b732f9e8dd7d2da52a8eb40f38de29c3b3d65afdb86f00/cfdonnx-2.0.1.tar.gz",
"platform": null,
"description": "# cfdonnx\n\nA Python module for exporting pre-trained CFD models to ONNX, making them interoperable with other ML frameworks and compatible with browsers.\n\nIt currently supports U-Net architecture and PyTorch models, but it will be soon extended to other frameworks and architectures.\n\nReproducible examples can be found at [openfoam-cfd-rom](https://github.com/simzero/openfoam-ml-rom) usign [DeepCFD](https://github.com/mdribeiro/DeepCFD).\n\n## Installation\n\nThe module can be installed with:\n\n```\npip3 install cfdonnx\n```\n\n## Usage\n\n```\nUsage: python3 -m cfdonnx [OPTIONS]\n\nOptions:\n -n, --net TEXT network architecture: UNetEx or AutoEncoder (default: UNetEx)\n -i, --input PATH checkpoint (default: checkpoint.pt)\n -o, --output PATH ONNX output file (default: checkpoint.onnx)\n -k, --kernel-size INT kernel size (optional, read from state_dict['kernel_size] by default )\n -f, --filters TEXT filter size, e.g. 8,16,32,32 (optional, read from state_dict['filters'] by default)\n -c --channels INT number of channels (optional, read from state_dict['input_shape'] by default)\n -x --nx INT X dimension (optional, read from state_dict['input_shape'] by default)\n -y --ny INT Y dimension (optional, read from state_dict['input_shape'] by default )\n -o, --output PATH Save model path (default: mymodel.pt)\n\nExample:\n\npython3 -m cfdonnx \\\n --net UNetEx \\\n --input flowAroundObstacles.pt \\\n --output flowAroundObstacles.onnx\n```\n\nYou can use your CFD ONNX models on runtime in Babylon.js as showcased at https://play.simzero.com/#D3SFTH#6 for the [flowAroundObstacles](https://github.com/simzero/openfoam-ml-rom/tree/main/OpenFOAM/incompressible/simpleFoam/flowAroundObstacles) example.\n\nA generic template for using ONNX is also available at https://play.simzero.com/#WIB297#1.\n\n\n",
"bugtrack_url": null,
"license": "MIT",
"summary": "Converting ML-CFD models to ONNX",
"version": "2.0.1",
"project_urls": {
"Homepage": "https://github.com/simzero/cfdonnx"
},
"split_keywords": [
"cfd",
"ml",
"onnx"
],
"urls": [
{
"comment_text": "",
"digests": {
"blake2b_256": "4ff3b2afcd5d89e6cb453c8fed9599bcde387ebcc9e32315b847aabc62f61c5f",
"md5": "3e229a44dc363eedcc7ca81b5af22ff9",
"sha256": "784690635f6fedb4f4a5bcf98b170f434a60ad5ffe7da4428dba94745fa32c73"
},
"downloads": -1,
"filename": "cfdonnx-2.0.1-py3-none-any.whl",
"has_sig": false,
"md5_digest": "3e229a44dc363eedcc7ca81b5af22ff9",
"packagetype": "bdist_wheel",
"python_version": "py3",
"requires_python": null,
"size": 9668,
"upload_time": "2023-07-04T21:16:44",
"upload_time_iso_8601": "2023-07-04T21:16:44.953653Z",
"url": "https://files.pythonhosted.org/packages/4f/f3/b2afcd5d89e6cb453c8fed9599bcde387ebcc9e32315b847aabc62f61c5f/cfdonnx-2.0.1-py3-none-any.whl",
"yanked": false,
"yanked_reason": null
},
{
"comment_text": "",
"digests": {
"blake2b_256": "db63da3c5c60e697c1b732f9e8dd7d2da52a8eb40f38de29c3b3d65afdb86f00",
"md5": "c46e66b8f34246c8da605f552d69c354",
"sha256": "c27302268541e9c1ffd703dd4b4e1b21ed93e3e453225be6843c48063d325216"
},
"downloads": -1,
"filename": "cfdonnx-2.0.1.tar.gz",
"has_sig": false,
"md5_digest": "c46e66b8f34246c8da605f552d69c354",
"packagetype": "sdist",
"python_version": "source",
"requires_python": null,
"size": 8355,
"upload_time": "2023-07-04T21:16:46",
"upload_time_iso_8601": "2023-07-04T21:16:46.519581Z",
"url": "https://files.pythonhosted.org/packages/db/63/da3c5c60e697c1b732f9e8dd7d2da52a8eb40f38de29c3b3d65afdb86f00/cfdonnx-2.0.1.tar.gz",
"yanked": false,
"yanked_reason": null
}
],
"upload_time": "2023-07-04 21:16:46",
"github": true,
"gitlab": false,
"bitbucket": false,
"codeberg": false,
"github_user": "simzero",
"github_project": "cfdonnx",
"travis_ci": false,
"coveralls": false,
"github_actions": false,
"requirements": [],
"lcname": "cfdonnx"
}
JSON
