Havuz

First open-source TurboQuant KV cache compression for LLM inference. Drop-in for HuggingFace. pip install turboquant.

Biologically-grounded adversarial training platform: cyclic Wake/Dream/Nightmare/Compress phases that accumulate model robustness without ca

Why is LLM inference slow — and how do you make it fast? A hands-on, first-principles course: roofline → KV cache → quantization → paralleli

A narrative walk through 90 years of AI history, paper by paper. 66 chapters from Turing 1936 to Blackwell 2025.

Recomputation-free, context-independent KV caching for large language models

End-to-end pipeline for seeing how LLMs actually process your prompts. Capture attention across every layer, render heatmaps and cooking cur

Model merging, task-vector rebasin, and fine-tuning for vision and LLM models.

Interactive, animated architecture maps for any Hugging Face model — no weights downloaded

A stateful AI Intrusion Prevention System (IPS) that reassembles TCP packet fragments in sequence order and uses PyTorch deep learning model

Build and train a mini Kimi K3 from scratch — a pure-PyTorch playground from a custom 88M model to 2T-scale LLM profiles.

ABTFarmer - Angry Birds Transformers farmer with automated time-skip glitch

Tiny-MoE is a lightweight Mixture-of-Experts language model built entirely from scratch in native PyTorch and trained end-to-end on Kaggle u

👻 MoEWatch: A lightweight diagnostic and audit library for Mixture-of-Experts (MoE) models in HuggingFace Transformers — the pytest for MoE

Companion code for the Manning book 'LLM Customization and Fine-Tuning.' Adapt open-weights LLMs end to end: prompting and RAG, LoRA/QLoRA,

Open benchmark for expert collapse and routing efficiency in sparse MoE LLMs

Models Take Notes at Prefill: KV Cache Can Be Editable and Composable (arXiv:2606.17107) — paper, code, results, and interactive companion.

本地 AI 驱动的实时对话翻译桌面应用（Real time dialogue translation desktop application driven by local AI）

A user's facts become a few rows in a content-addressed memory table — not a rewrite of the model. Paper, code, data, and interactive site f

A free, open GenAI curriculum built from scratch. Start from zero, build real projects, and work your way up to production AI systems.

Seven Minutes Is All I Can Spare To Play With You.

Local-first social memory search engine with browser capture, hybrid AI retrieval, and optional C++ acceleration.

Mechanistic experiments on how LLMs represent and compute arithmetic internally, with strict no-parser controls, reproducible audits, and an

Train decoder-only language models from scratch with PyTorch, featuring RoPE, RMSNorm, SwiGLU, and SDPA

Desktop app for browsing and filtering the Xbox Game Pass catalog with on-device AI genre classification