- The Rest Is Silence of Code
Starred repositories
prime (previously called ZeroBand) is a framework for efficient, globally distributed training of AI models over the internet.
CIEL Is an Extended Lisp. Scripting with batteries included.
A Ninja's Handbook: A book on privacy, security, and anonymity online.
Flux diffusion model implementation using quantized fp8 matmul & remaining layers use faster half precision accumulate, which is ~2x faster on consumer devices.
Uses LangChain, gpt-4 and gpt-3.5-turbo to write an initial prompt, then iteratively (automagically) perfect it by testing against a small dataset until high accuracy is achieved.
Boosting your Web Services of Deep Learning Applications.
ReFT: Representation Finetuning for Language Models
Tiny CPU is a small 32-bit CPU done mostly as a hobby for educational purposes.
A pytorch quantization backend for optimum
This repository consists of documentation regarding quantum machine learning (QML), covering both beginner's guides and advanced topics.
Official PyTorch Implementation of "The Hidden Attention of Mamba Models"
Official implementation of Half-Quadratic Quantization (HQQ)
The open-source Roblox & UEFN alternative giving you freedom to own what you create. An all-in-one, real-time, collaborative game development platform built on Godot.
Summarize existing representative LLMs text datasets.
A functional, flexible and concise Lisp.
Freeing data processing from scripting madness by providing a set of platform-agnostic customizable pipeline processing blocks.
Computer science and computer-adjacent papers (and sometimes books) that have influenced me deeply.
Finetune Llama 3.2, Mistral, Phi & Gemma LLMs 2-5x faster with 80% less memory
The most powerful and modular diffusion model GUI, api and backend with a graph/nodes interface.
Simple and efficient pytorch-native transformer text generation in <1000 LOC of python.