# 🧠 Nanograd - Lightweight Autograd Engine for Deep Learning
**Nanograd** is a minimalistic automatic differentiation engine for building and training neural networks. Inspired by PyTorch’s autograd, it provides an easy-to-use framework for defining computational graphs, performing backpropagation, and training models.
## 🚀 Features
✅ **Automatic Differentiation** - Compute gradients with ease using backpropagation.
✅ **Graph-Based Computation** - Uses a dynamic computation graph to track operations.
✅ **Lightweight & Fast** - No unnecessary dependencies, optimized for speed.
✅ **Custom Neural Networks** - Build and train models from scratch.
✅ **Graph Visualization** - Visualize computational graphs using `graphviz`.
---
## 📦 Installation
You can install **Nanograd** directly from PyPI:
```bash
pip install nanograd
```
## 🔧 Usage
### 1️⃣ Defining Computation
```python
from nanograd.engine import Value
a = Value(2.0)
b = Value(3.0)
c = a * b + 5
c.backward()
print(f"Value of c: {c.data}") # Output: 11.0
print(f"Gradient of a: {a.grad}") # Output: 3.0
print(f"Gradient of b: {b.grad}") # Output: 2.0
```
### 2️⃣ Building a Neural Network
```python
from nanograd.nn import MLP
import numpy as np
# Create a 2-layer neural network (2 inputs, 4 hidden, 1 output)
model = MLP(2, [4, 1])
# Dummy data
X = np.array([[1.0, 2.0]])
y = np.array([1.0])
# Forward pass
pred = model.forward(X)
print(f"Prediction: {pred}")
```
### 3️⃣ Visualizing Computational Graph
```python
from nanograd.graph import draw_graph
draw_graph(c) # Generates a graph of computations
```
Raw data
{
"_id": null,
"home_page": "https://github.com/NeuralNoble/nanograd",
"name": "nanograd-aman",
"maintainer": null,
"docs_url": null,
"requires_python": ">=3.6",
"maintainer_email": null,
"keywords": null,
"author": "Aman Anand",
"author_email": "coursesxyz403@gmail.com",
"download_url": "https://files.pythonhosted.org/packages/e3/5b/2468ef9f48111e2821bffbd1f60a7d336fab2e8d492036511e05073ab6eb/nanograd-aman-0.1.0.tar.gz",
"platform": null,
"description": "# \ud83e\udde0 Nanograd - Lightweight Autograd Engine for Deep Learning \n\n**Nanograd** is a minimalistic automatic differentiation engine for building and training neural networks. Inspired by PyTorch\u2019s autograd, it provides an easy-to-use framework for defining computational graphs, performing backpropagation, and training models. \n\n## \ud83d\ude80 Features \n\u2705 **Automatic Differentiation** - Compute gradients with ease using backpropagation. \n\u2705 **Graph-Based Computation** - Uses a dynamic computation graph to track operations. \n\u2705 **Lightweight & Fast** - No unnecessary dependencies, optimized for speed. \n\u2705 **Custom Neural Networks** - Build and train models from scratch. \n\u2705 **Graph Visualization** - Visualize computational graphs using `graphviz`. \n\n---\n\n## \ud83d\udce6 Installation \n\nYou can install **Nanograd** directly from PyPI: \n\n```bash\npip install nanograd\n```\n\n\n## \ud83d\udd27 Usage\n\n### 1\ufe0f\u20e3 Defining Computation\n\n```python\nfrom nanograd.engine import Value\n\na = Value(2.0)\nb = Value(3.0)\nc = a * b + 5\nc.backward()\n\nprint(f\"Value of c: {c.data}\") # Output: 11.0\nprint(f\"Gradient of a: {a.grad}\") # Output: 3.0\nprint(f\"Gradient of b: {b.grad}\") # Output: 2.0\n\n```\n\n### 2\ufe0f\u20e3 Building a Neural Network\n```python\nfrom nanograd.nn import MLP\nimport numpy as np\n\n# Create a 2-layer neural network (2 inputs, 4 hidden, 1 output)\nmodel = MLP(2, [4, 1])\n\n# Dummy data\nX = np.array([[1.0, 2.0]])\ny = np.array([1.0])\n\n# Forward pass\npred = model.forward(X)\nprint(f\"Prediction: {pred}\")\n\n```\n\n### 3\ufe0f\u20e3 Visualizing Computational Graph\n\n```python\nfrom nanograd.graph import draw_graph\ndraw_graph(c) # Generates a graph of computations\n\n```\n\n\n",
"bugtrack_url": null,
"license": null,
"summary": "A lightweight backpropagation package for neural networks",
"version": "0.1.0",
"project_urls": {
"Homepage": "https://github.com/NeuralNoble/nanograd"
},
"split_keywords": [],
"urls": [
{
"comment_text": null,
"digests": {
"blake2b_256": "e35b2468ef9f48111e2821bffbd1f60a7d336fab2e8d492036511e05073ab6eb",
"md5": "50286f4a9a538f0ef0c9384fb904d9ba",
"sha256": "d44b5e9936f74fde791ad2149d0d5d0fff1d41a4fa9aacc70a89ec1e3c6d2b33"
},
"downloads": -1,
"filename": "nanograd-aman-0.1.0.tar.gz",
"has_sig": false,
"md5_digest": "50286f4a9a538f0ef0c9384fb904d9ba",
"packagetype": "sdist",
"python_version": "source",
"requires_python": ">=3.6",
"size": 3896,
"upload_time": "2025-01-31T18:10:41",
"upload_time_iso_8601": "2025-01-31T18:10:41.321920Z",
"url": "https://files.pythonhosted.org/packages/e3/5b/2468ef9f48111e2821bffbd1f60a7d336fab2e8d492036511e05073ab6eb/nanograd-aman-0.1.0.tar.gz",
"yanked": false,
"yanked_reason": null
}
],
"upload_time": "2025-01-31 18:10:41",
"github": true,
"gitlab": false,
"bitbucket": false,
"codeberg": false,
"github_user": "NeuralNoble",
"github_project": "nanograd",
"travis_ci": false,
"coveralls": false,
"github_actions": false,
"lcname": "nanograd-aman"
}
JSON
