arclm 0.1.0

A compact PyTorch language-model training and fine-tuning library.

ArcLM

ArcLM is a lightweight PyTorch library for training, fine-tuning, and generating text with compact causal language models.

It is designed for learning, experimentation, research, and local AI workflows.

Features

• Train compact causal language models from your own text
• Generate text with a simple API
• Fine-tune existing ArcLM checkpoints
• Lightweight and easy to understand
• Built on PyTorch

Installation

Install ArcLM from PyPI:

pip install arclm

Or install the latest development version:

git clone https://github.com/Ahmad-Al-Dibo/ArcLM.git
cd ArcLM
pip install -e ".[dev]"

Quick Start

Train a model:

python train.py

Generate text:

from arclm import load_model

model = load_model("models/arclm.pth")
print(model.predict("machine learning", max_new_tokens=50))

Documentation

The complete documentation is available in docs/USAGE.md.

It includes:

• Training a model from scratch
• Loading saved checkpoints
• Fine-tuning existing ArcLM models
• Working with external checkpoints
• Configuration options
• Best practices and notes

Project Structure

arclm/              Library source code
data/               Example datasets
docs/               Documentation
examples/           Example scripts
models/             Local checkpoints
tests/              Tests

License

ArcLM is licensed under the Apache License 2.0. See the LICENSE file for details.