🌱 LittleLearn – Touch the Big World with Little Steps

update Version (1.2.1) date : (12-February-2026):

- add Doc String  
- add new Optimizer NAdam,SGD,Adagrad and Momentum

warning : on this update we remove so many feature because paradims changed.

LittleLearn is an experimental and original machine learning framework built from scratch — inspired by the simplicity of Keras and the flexibility of PyTorch, yet designed with its own architecture, philosophy, and gradient engine.

🧠 What Makes LittleLearn Different?

🔧 Not a wrapper – LittleLearn is not built on top of TensorFlow, PyTorch, or other major ML libraries.
💡 Fully original layers, modules, and autodiff engine (GradientReflector).
🧩 Customizable down to the node level: build models from high-level APIs or go low-level for complete control.
🛠️ Features unique like:
Node-level gradient clipping
Inline graph tracing
Custom attention mechanisms (e.g., Multi-Head Attention from scratch)
🤯 Designed for both research experimentation and deep learning education.

⚙️ Core Philosophy

Touch the Big World with Little Steps. Whether you want rapid prototyping or total model control — LittleLearn gives you both.

📦 Ecosystem Features

✅ Deep learning modules: Dense, LSTM, attention mechanisms, and more
🤖 instant model by Model Module
🔄 Custom training loops with full backend access
🧠 All powered by the GradientReflector engine — providing automatic differentiation with transparency and tweakability

🔧 Installation

    pip install littlelearn

🚀 Quick Example :

    import littlelearn as ll 
    import littlelearn.DeepLearning as dl 

    model = dl.layers.Sequential([
        dl.layers.Linear(20,32),
        dl.activations.Relu(),
        dl.layers.Linear(32,1)
    ]) 
    model.train()
    x_train,y_train= datasets()
    optimizer = dl.optimizers.Adam(model.parameter())

    for epoch in range(100) :
        y_pred = model(x_train)
        loss = dl.loss.mse_loss(y_train,y_pred)
        loss.backwardpass()
        optimizer.step()
        loss.reset_grad()
        print(loss.tensor)
    
    model.inference()
    model.save("model.npz")

📌 Disclaimer While inspired by well-known frameworks, LittleLearn is built entirely from scratch with its own mechanics. It is suitable for:

🔬 Experimental research
🏗️ Framework building
📚 Educational purposes
🔧 Custom low-level operations

suport this project : https://ko-fi.com/alpin92578

👤 Author Candra Alpin Gunawan 📧 hinamatsuriairin@gmail.com 🌐 GitHub https://github.com/Airinchan818/LittleLearn

youtube : https://youtube.com/@hinamatsuriairin4596?si=KrBtOhXoVYnbBlpY

Metadata

Release files for littlelearn 1.2.3

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

Source distribution (sdist)

Source distribution for littlelearn 1.2.3
File	Size	Uploaded
littlelearn-1.2.3.tar.gz	51.3 kB	Feb 14, 2026	Details

Built distribution (wheel)

Table of built distributions (wheels) for littlelearn 1.2.3
File	Interpreter	ABI	Platform	Reset
littlelearn-1.2.3-py3-none-any.whl 54.2 kB Feb 14, 2026	Python 3	none	any	Details

Total release size: 105.5 kB

Release files / littlelearn-1.2.3.tar.gz

Download URL	littlelearn-1.2.3.tar.gz
Size	51.3 kB
Tags	Source
SHA-256 checksum How to use checksums	`de36e873312bd5cacef88f950dfa3ae8e2fdc384cfaa081c5664fa186a321db3`
BLAKE2b-256 checksum How to use checksums	`f77d0abdd52114c24d74e56f2f875414994a79ce30f84aac1dc72857648e01bb`
Upload date	Feb 14, 2026
Uploaded using Trusted Publishing? What is trusted publishing?	No
Uploaded via	`twine/6.1.0 CPython/3.12.7`

Release files / littlelearn-1.2.3-py3-none-any.whl

Download URL	littlelearn-1.2.3-py3-none-any.whl
Size	54.2 kB
Tags	Python 3
SHA-256 checksum How to use checksums	`1b57b3f901ad26281c342d341ab4ef57e29ab959fc61901da98ad106acfcaf58`
BLAKE2b-256 checksum How to use checksums	`67454c67558cd13d5939b6d07f8e0bf449dab0dcb0e11af334ea045ba964cf7d`
Upload date	Feb 14, 2026
Uploaded using Trusted Publishing? What is trusted publishing?	No
Uploaded via	`twine/6.1.0 CPython/3.12.7`