26m function call model that runs on incredibly small devices

A lightweight transformer language model built from scratch in PyTorch, trained on a single consumer GPU with a full pipeline for data processing, pretraining, and instruction tuning.

1bit llama.cpp gguf weights paired with turboquant 4 bit kc cache

I replicated Ng's RYS method and found that duplicating 3 specific layers in Qwen2.5-32B boosts reasoning by 17% and duplicating layers 12-14 in Devstral-24B improves logical deduction from 0.22→0.…

ASCII art via Gradient Descent

PersonaLive! : Expressive Portrait Image Animation for Live Streaming

Collection of extracted System Prompts from popular chatbots like ChatGPT, Claude & Gemini

State-of-the-art TTS model under 25MB 😻

Quickly render fractals in CUDA