Skip to main content

🚀 GradLite

A tiny, elegant automatic differentiation engine built from scratch.

GradLite is a lightweight module for automatic gradient computation, the core mechanism behind training neural networks using backpropagation.

Inspired by:

...but extended to support more features than MicroGrad, without all the complexity of PyTorch. This module supports building complete neural networks and has a clear educational purpose.


✨ Features

  • 🧠 Reverse-mode automatic differentiation (backpropagation)
  • 🔢 Scalar-based computational graph
  • 🔄 Dynamic graph construction
  • 🏗️ Build full neural networks (Linear layer support for now! 🗓️)
  • 🪶 Lightweight and easy to read
  • 📚 Fully Educational

📦 Installation

You can run the following command to install gradlite in your system/virtual environment:

pip install gradlite

🧩 Example Usage

The usage of gradlite to build expressions and run backpropagation (automatic differentation) is shown next:

from gradlite.core import Parameter

a = Parameter(2.0)
b = Parameter(0.3)
c = Parameter(1.0)

ab = a * b
f = ab + c
print(f'f={f.value:.4f}')
f.backward()
print(f'grad_f={f.grad:.4f}')
print(f'grad_a={a.grad:.4f}')
print(f'grad_b={b.grad:.4f}')
print(f'grad_c={c.grad:.4f}')

An example training loop of gradlite is given next:

from gradlite.nn.linear import Linear
from gradlite.nn.activations import ReLU
from gradlite.nn.loss.mse import MSE
from gradlite.optimizer.sgd import SGD

X = [2.00, 2.10, 2.20, 2.30, 2.40]
y = [3.00, 3.10, 3.13, 3.16, 3.19]

epochs = 2000
model = Linear(input_features=1, output_features=1, activation_fn=ReLU())
model.neurons[0].weights[0].value = 0.5  # Using positive weight to prevent Dying ReLU
optimizer = SGD(model.parameters(), lr=1e-2)
loss_fn = MSE()

for _ in range(epochs):
    for x_value, y_true in zip(X, y):
        optimizer.zero_grad()
        y_pred = model([x_value])
        loss = loss_fn(y_pred, [y_true])
        loss.backward()
        optimizer.step()
        print(f'Loss: {loss.value:.4f}')

Metadata

Release files for gradlite 0.2.1

For a detailed explanation of source distributions (sdists) and built distributions (wheels), please see the package formats documentation.

Source distribution (sdist)

Source distribution for gradlite 0.2.1
File Size Uploaded Provenance Actions
gradlite-0.2.1.tar.gz 117.7 kB Without provenance Details for gradlite-0.2.1.tar.gz

Built distribution (wheel)

Table of built distributions (wheels) for gradlite 0.2.1
File Interpreter ABI Platform Actions
gradlite-0.2.1-py3-none-any.whl Python 3 none any Details for gradlite-0.2.1-py3-none-any.whl

Total release size: 129.2 kB

Release files / gradlite-0.2.1.tar.gz

Download URL gradlite-0.2.1.tar.gz
Size 117.7 kB
Tags Source
SHA-256 checksum
How to use checksums
3f6f7019ce61f78f3a812140fa1745785eef199e539290a27f15d1358d52912b
BLAKE2b-256 checksum
How to use checksums
1a0c02dcafbb3a3bb0873832829a8d7cc322bfdbfa583c2e7585928fc909ce8e
Upload date
Uploaded using Trusted Publishing?
What is trusted publishing?
No
Uploaded via twine/6.2.0 CPython/3.12.3

Release files / gradlite-0.2.1-py3-none-any.whl

Download URL gradlite-0.2.1-py3-none-any.whl
Size 11.5 kB
Tags Python 3
SHA-256 checksum
How to use checksums
3368f81288af2ade12437ef63b59a3f250a4d5b59a8410af4948614f22bd60b0
BLAKE2b-256 checksum
How to use checksums
14389852641c18091a3187fb8d7e8c5be9c097bfbe013961d0c5580c3020278c
Upload date
Uploaded using Trusted Publishing?
What is trusted publishing?
No
Uploaded via twine/6.2.0 CPython/3.12.3

Security for gradlite 0.2.1

Provenance

Provenance describes where a file came from. On PyPI, provenance is shared via attestations, which provide a verifiable record of the build or publishing details. View details, limitations and caveats.

No provenance found

None of the files in this release were uploaded with provenance metadata.

A note on package security

PyPI is a repository for distributing packages, not a service for auditing them. While third-party security researchers and our staff do their best to detect and remove malicious packages, PyPI cannot guarantee the quality, security or suitability of any given package.

To learn more about protecting yourself from malicious actors, visit our package security documentation.

Release history Release notifications | RSS feed

This release

0.2.1 This release

2 release files No provenance

Anthropic, PBC Visionary sponsor Bloomberg Visionary sponsor Hudson River Trading Visionary sponsor Meta Visionary sponsor NVIDIA Visionary sponsor Microsoft Sustainability sponsor Depot Continuous Integration AWS Cloud computing and Security Sponsor Datadog Monitoring Fastly CDN Google Download Analytics Sentry Error logging StatusPage Status page