arithmetic-coding

Entropy coding / arithmetic coding for PyTorch

Data Compression using Arithmetic Encoding in Python

The implementation of various data compression techniques.

Nacrith — Lossless text compression via ensemble neural arithmetic coding. Combines SmolLM2-135M language model with context mixing, adaptive prediction, and high-precision CDF coding. 3.1x better than gzip, and outperforming CMIX, ts_zip, LLMZip and FineZip. Fully lossless. GPU accelerated.

LLMA = LLM + Arithmetic coder, which use LLM to do insane text data compression. LLMA=大模型+算术编码，它能使用LLM对文本数据进行暴力的压缩，达到极高的压缩率。

Arithmetic coding library with statistical models like PPM and Context mixing (demonstrating core principles of probabilistic inference, ensemble learning in AI)

Image compression using arithmetic coding algorithm

Autonomous Agent 32-bit Finite-Precision Arithmetic Coding Lossless Compression Engine with Underflow Prevention

Autonomous Agent 32-bit Finite-Precision Arithmetic Coding Lossless Compression Engine with Underflow Prevention

High-precision fractional range subdivision arithmetic compression encoder and decoder

High-precision fractional range subdivision arithmetic compression encoder and decoder

An accurate language model is a high-compression, lossless data compressor

Implementation of lossless data compression technique written in Python.

Steganographic data encoding in LLM-generated text via arithmetic coding against the model's own next-token distribution.

implementation of the CTW algorithm

My own Unicode compression algorithm

🗃 Implementation of encoding and decoding of Arithmetic-Coding algorithm using python..

Experiment applying L3C (CVPR2019) for image compression on CLIC-2019 dataset

LLM-Enhanced Text Compression: This project explores using Large Language Models (LLMs) to improve text compression efficiency. It evaluates various LLM architectures (GPT-2, Llama 2, Mistral-7B, Yi-34B) across different context sizes and datasets, implementing a rank-based compression method integrated with conventional algorithms.