[](https://www.python.org/downloads/release/python-360/)
[](https://pypi.org/project/torchy-nn/)
## Overview
Torchy is a neural network framework implemented only using NumPy and based on PyTorch API but with manual backpropogation calculations. The main idea was to build a neural network from scratch for educational purposes.
## Installation
```python
pip install torchy-nn
```
## Getting started
I suggest you to take a look at [currently implemented stuff](https://github.com/chuvalniy/Torchy/blob/main/docs/Implemented.md) to be familiar with current possibilities for building neural network models with Torchy. Also I've created [package structure](https://github.com/chuvalniy/Torchy/blob/main/docs/PackageStructure.md) in case if you stuck where to get specific layers.
## Example usage
First we can define our model using Torchy with its PyTorch-like API
```python
from nn.modules.sequential import Sequential # Same as nn.Sequential
import nn.modules.module as layer
# Define 2-layer wtth 100 neurons hidden layer.
model = Sequential(
layer.Linear(n_input=10, n_output=100),
layer.BatchNorm1d(n_output=100),
layer.ReLU(),
layer.Linear(n_input=100, n_output=2)
)
```
Next step is to create instances of optimizer and criterion for loss function and scheduler for fun
```python
import nn.modules.loss as loss
import optimizers.optim as optim
import optimizers.scheduler as sched
optimizer = optim.SGD(model.params(), lr=1e-3)
criterion = loss.CrossEntropyLoss()
scheduler = sched.StepLR(optimizer, step_size=10)
```
I won't cover whole training process like loops and stuff, just show you main differences while training
```python
...
predictions = model(X) # Nothing changed
loss, grad = criterion(predictions, y) # Now return tuple of (loss, grad) instead of only loss
optimizer.zero_grad()
model.backward(grad) # Call backward on model object and pass gradient from loss as argument
optimizer.forward_step()
```
## Demonstration
The [demo notebook](https://github.com/chuvalniy/Torchy/blob/main/torchy-demo.ipynb) showcases what Torchy currently can do.
## Roadmap
There is still a lot of work to be done, but here are the main points that will be completed soon
- Docstring every entity & add type hinting
- Add evaluation & inference for model
## Resources
The opportunity to create such a project was given to me thanks to these people
- [PyTorch](https://github.com/pytorch/pytorch)
- [CS231n - 2016](https://youtube.com/playlist?list=PLkt2uSq6rBVctENoVBg1TpCC7OQi31AlC)
- [Deep Learning на пальцах - 2019](https://youtube.com/playlist?list=PL5FkQ0AF9O_o2Eb5Qn8pwCDg7TniyV1Wb)
- [Neural Networks: Zero to Hero](https://youtube.com/playlist?list=PLAqhIrjkxbuWI23v9cThsA9GvCAUhRvKZ)
## License
[MIT License](https://github.com/chuvalniy/Torchy/blob/main/LICENSE)
Raw data
{
"_id": null,
"home_page": "https://github.com/chuvalniy/Torchy",
"name": "torchy-nn",
"maintainer": "",
"docs_url": null,
"requires_python": "",
"maintainer_email": "",
"keywords": "python,neural net,from scratch,numpy,pytorch-like,cnn,dense",
"author": "Valentin Belyaev",
"author_email": "chuvalik.work@gmail.com",
"download_url": "https://files.pythonhosted.org/packages/2c/b4/5ead303ceff61764eb22a27cc594720288856ee8cc7efa78a190f84fcfe7/torchy-nn-0.3.5.3.tar.gz",
"platform": null,
"description": "\n[](https://www.python.org/downloads/release/python-360/)\n[](https://pypi.org/project/torchy-nn/)\n\n\n\n\n\n## Overview\nTorchy is a neural network framework implemented only using NumPy and based on PyTorch API but with manual backpropogation calculations. The main idea was to build a neural network from scratch for educational purposes.\n\n## Installation\n```python\npip install torchy-nn\n```\n## Getting started\nI suggest you to take a look at [currently implemented stuff](https://github.com/chuvalniy/Torchy/blob/main/docs/Implemented.md) to be familiar with current possibilities for building neural network models with Torchy. Also I've created [package structure](https://github.com/chuvalniy/Torchy/blob/main/docs/PackageStructure.md) in case if you stuck where to get specific layers.\n\n## Example usage\nFirst we can define our model using Torchy with its PyTorch-like API\n\n```python\nfrom nn.modules.sequential import Sequential # Same as nn.Sequential\nimport nn.modules.module as layer\n\n# Define 2-layer wtth 100 neurons hidden layer.\nmodel = Sequential(\n layer.Linear(n_input=10, n_output=100),\n layer.BatchNorm1d(n_output=100),\n layer.ReLU(),\n layer.Linear(n_input=100, n_output=2)\n)\n```\n\nNext step is to create instances of optimizer and criterion for loss function and scheduler for fun\n\n```python\nimport nn.modules.loss as loss\nimport optimizers.optim as optim\nimport optimizers.scheduler as sched\n\noptimizer = optim.SGD(model.params(), lr=1e-3)\ncriterion = loss.CrossEntropyLoss()\nscheduler = sched.StepLR(optimizer, step_size=10)\n```\n\nI won't cover whole training process like loops and stuff, just show you main differences while training\n\n```python\n...\npredictions = model(X) # Nothing changed\n\nloss, grad = criterion(predictions, y) # Now return tuple of (loss, grad) instead of only loss \n\noptimizer.zero_grad()\nmodel.backward(grad) # Call backward on model object and pass gradient from loss as argument\noptimizer.forward_step()\n```\n\n\n## Demonstration\nThe [demo notebook](https://github.com/chuvalniy/Torchy/blob/main/torchy-demo.ipynb) showcases what Torchy currently can do.\n\n## Roadmap\nThere is still a lot of work to be done, but here are the main points that will be completed soon\n- Docstring every entity & add type hinting\n- Add evaluation & inference for model \n\n## Resources\nThe opportunity to create such a project was given to me thanks to these people\n\n- [PyTorch](https://github.com/pytorch/pytorch)\n- [CS231n - 2016](https://youtube.com/playlist?list=PLkt2uSq6rBVctENoVBg1TpCC7OQi31AlC)\n- [Deep Learning \u043d\u0430 \u043f\u0430\u043b\u044c\u0446\u0430\u0445 - 2019](https://youtube.com/playlist?list=PL5FkQ0AF9O_o2Eb5Qn8pwCDg7TniyV1Wb)\n- [Neural Networks: Zero to Hero](https://youtube.com/playlist?list=PLAqhIrjkxbuWI23v9cThsA9GvCAUhRvKZ)\n\n\n## License\n[MIT License](https://github.com/chuvalniy/Torchy/blob/main/LICENSE)\n\n",
"bugtrack_url": null,
"license": "MIT",
"summary": "NumPy based neural network package with PyTorch-like API",
"version": "0.3.5.3",
"project_urls": {
"Homepage": "https://github.com/chuvalniy/Torchy"
},
"split_keywords": [
"python",
"neural net",
"from scratch",
"numpy",
"pytorch-like",
"cnn",
"dense"
],
"urls": [
{
"comment_text": "",
"digests": {
"blake2b_256": "0ac3cae1c93872245c04c78cd60f289c2ae199366c3ccfc280bd9b681fba4a9c",
"md5": "b2c56c467a65d73f342618f74b460a14",
"sha256": "f394083044302df1c5fa939fd6c96552bb608e55ab6b030655d27a5a4dee1e88"
},
"downloads": -1,
"filename": "torchy_nn-0.3.5.3-py3-none-any.whl",
"has_sig": false,
"md5_digest": "b2c56c467a65d73f342618f74b460a14",
"packagetype": "bdist_wheel",
"python_version": "py3",
"requires_python": null,
"size": 28757,
"upload_time": "2023-10-19T17:11:10",
"upload_time_iso_8601": "2023-10-19T17:11:10.390979Z",
"url": "https://files.pythonhosted.org/packages/0a/c3/cae1c93872245c04c78cd60f289c2ae199366c3ccfc280bd9b681fba4a9c/torchy_nn-0.3.5.3-py3-none-any.whl",
"yanked": false,
"yanked_reason": null
},
{
"comment_text": "",
"digests": {
"blake2b_256": "2cb45ead303ceff61764eb22a27cc594720288856ee8cc7efa78a190f84fcfe7",
"md5": "8f1d1ec6d0b863f8c711f5301b3cce8a",
"sha256": "5ebf9e9946698a10f62e542d95b050d040be34c2063e5023a09aad974d3b622c"
},
"downloads": -1,
"filename": "torchy-nn-0.3.5.3.tar.gz",
"has_sig": false,
"md5_digest": "8f1d1ec6d0b863f8c711f5301b3cce8a",
"packagetype": "sdist",
"python_version": "source",
"requires_python": null,
"size": 15415,
"upload_time": "2023-10-19T17:11:12",
"upload_time_iso_8601": "2023-10-19T17:11:12.638782Z",
"url": "https://files.pythonhosted.org/packages/2c/b4/5ead303ceff61764eb22a27cc594720288856ee8cc7efa78a190f84fcfe7/torchy-nn-0.3.5.3.tar.gz",
"yanked": false,
"yanked_reason": null
}
],
"upload_time": "2023-10-19 17:11:12",
"github": true,
"gitlab": false,
"bitbucket": false,
"codeberg": false,
"github_user": "chuvalniy",
"github_project": "Torchy",
"travis_ci": false,
"coveralls": false,
"github_actions": false,
"lcname": "torchy-nn"
}
JSON
