nanoautograd

nanoautograd is a lightweight Python library that provides a minimalistic automatic differentiation engine, inspired by PyTorch's autograd system. It allows for the construction of computational graphs and automatic computation of gradients via backpropagation. It supports basic tensor operations and activation functions.

Installation

1. Clone the repository:

   git clone https://github.com/sindhu213/nanoautograd.git
   cd nanoautograd

2. (Optional) Create and activate a virtual environment:

   python -m venv venv
   source venv/bin/activate  # On Windows: venv\Scripts\activate

3. Install the required dependencies:

   pip install -r requirements.txt

Usage

from nanoautograd.tensor import Tensor

# Create input tensors
a = Tensor(2.0, requires_grad=True)
b = Tensor(3.0, requires_grad=True)

# Perform operations
c = a * b
d = c + a

# Compute gradients
d.backward()

# Access gradients
print(f"Gradient of a: {a.grad}")
print(f"Gradient of b: {b.grad}")

Contributing

Contributions are welcome! If you'd like to contribute, please fork the repository and submit a pull request.