AIscripts

AIscripts is a collection of experimental algorithms, machine learning components, and optimization tools inspired by various domains. This repository serves as a playground for exploring computational efficiency, mathematical models, and AI-driven solutions.

---

Features

Core Components

• Matrix Operations:

  GEMM_tiled.py - Tiled implementation of General Matrix Multiply (GEMM).

  KMM.py - Karatsuba Matrix Multiplication utilities.

  MGD.py - Meta Gradient Descent optimizer.

• AI/ML Tools:

  SeedLM.py - Seed and coefficient selection algorithm for a single weight block.

  aicommit.py - AI-assisted Git commit message generator.

  barcode_ssd_mobilenet_v1_dmp25_quant.tflite.runner.py - TensorFlow Lite runner for SSD MobileNet-based barcode detection.

  ZeroMerge.py - Parameter-Free KV Cache Compression for Memory-Efficient Long-Context LLMs.

  NoProp.py - Training Neural Networks without Back-propagation or Forward-propagation.

  PhyloLM.ipynb and phylolm.py - Inferring the Phylogeny of Large Language Models and Predicting their Performances in Benchmarks.

  DLFloat.py - Lossless LLM Compression for Efficient GPU Inference via Dynamic-Length Float

• Probabilistic Data Structures:

  bloomfilter.py - Classic Bloom filter implementation.

  cuckoofilter.py - Space-efficient Cuckoo filter.

• Optimization & Cryptography:

  minimal_feistel_network.py - Compact Feistel network cipher.

• Combinatorial Algorithms:

  polyomino_tiling.py - Polyomino tiling solver.

  cassowary.py - Constraint-solving algorithm implementation.

• Misc:

  calcDa.py - Density altitude calculator.

  mdnscan.py - mdns simple scanner.

  mbf.py - Minimal bloom filter implementation.

---

Getting Started

Prerequisites

• Python 3.8+
• pip install -r requirements.txt
  (Example dependencies: NumPy, TensorFlow Lite, PyTorch)

Basic Usage

# Run matrix multiplication benchmark
python GEMM_tiled.py --size 1024

# Generate AI-assisted commit message
python aicommit.py --diff <your_git_diff>