Alysium TinyLM

Lightweight • Pure NumPy • Educational

A tiny character-free, pure NumPy language model built for learning and experimentation.

What is Alysium TinyLM?

Alysium TinyLM is a small educational language model written entirely in pure NumPy.
It was created to help people understand how language models work under the hood — without heavy frameworks like PyTorch or TensorFlow.

Word-level next-word prediction
Extremely lightweight
Easy to train on your own data
Simple chat interface
Perfect for learning and experimentation

Features

Pure NumPy implementation (no PyTorch / TensorFlow)
Easy to understand and modify
Train on your own dataset.txt
Interactive chat mode
Small enough to run on limited hardware

Quick Start

# Clone the repository
git clone https://github.com/alysiumcorporationstudios/AlysiumTinyLLM.git
cd AlysiumTinyLLM

# Install dependency
pip install numpy

# Chat with the model
python chat.py

Training your own model

Edit dataset.txt with your own conversations (one utterance per line)
Run training:

python train.py

After training finishes, a new tinylm.pkl will be created.
Run python chat.py to talk to your updated model.

Project Structure

AlysiumTinyLLM/
├── model.py          # The neural network
├── train.py          # Training script
├── chat.py           # Interactive chat
├── config.py         # Model & training settings
├── dataset.txt       # Training data
├── tinylm.pkl        # Trained model weights
└── README.md

Configuration

You can change model size and training settings in config.py:

EMBED_DIM = 64
HIDDEN_DIM = 128
CONTEXT_SIZE = 5
EPOCHS = 25
LEARNING_RATE = 0.025

Example

You: hi
AI:  hello how are you.

You: what is your name
AI:  my name is alysium.

You: bye
AI:  goodbye.

Socials

X (Twitter): @AlysiumCorpZA
Instagram: @iam_realtoxicdeemodder

License

This project is open source under the MIT License.

Made with ❤️ by Alysium Corporation Studios

Downloads last month: -; Downloads are not tracked for this model. How to track