The definitive list of the best of (consumer facing) open source.
Class activation maps for your PyTorch models (CAM, Grad-CAM, Grad-CAM++, Smooth Grad-CAM++, Score-CAM, SS-CAM, IS-CAM, XGrad-CAM, Layer-CAM, Finer-CAM, LeGrad, RefineCAM)
Master the fundamentals of machine learning, deep learning, and mathematical optimization by building key concepts and models from scratch using Python.
Tweak the form field rendering in templates, not in python-level form definitions. CSS classes and HTML attributes can be altered.
A comprehensive list of papers for the definition of World Models and using World Models for General Video Generation, Embodied AI, and Autonomous Driving, including papers, codes, and related websites.
π½ Build Bespoke OS Images
Python Materials Genomics (pymatgen) is a robust materials analysis code that defines classes for structures and molecules with support for many electronic structure codes. It powers the Materials Project.
MetricFlow allows you to define, build, and maintain metrics in code.
Python3 library for downloading YouTube Videos.
A clean, self-hosted web UI for managing your Traefik reverse proxy.
Open specification for defining and expressing service level objectives (SLO)
OpenCL integration for Python, plus shiny features
This library reads and writes ESRI Shapefiles in pure Python.
η§εΉ»ε°θ―΄δ½ει
Docker Swarm mode rocks! Ideas, tools and recipes. Get a production-ready, distributed, HTTPS served, cluster in minutes, not weeks.
Collection of various BIOS/UEFI-related utilities which aid in research and/or modding purposes.
Correctly generate plurals, ordinals, indefinite articles; convert numbers to words
π¦ The unofficial Python client for the Uniswap exchange.
A Python library to access ISO country, subdivision, language, currency and script definitions and their translations.
Easy access to official spatial data sets of Brazil in R and Python
Parse BIOS/Intel ME/UEFI firmware related structures: Volumes, FileSystems, Files, etc
vMLX - Use MLX models easily - JANGQ (GGUF for MLX) - Not dependant on mlx_vlm
High Octane Triage Analysis
A Python library for trading automation on DeFi, data research and integration. Supporting Uniswap, Aave, Chainlink, USDC and other protocols.
Machine learning for multivariate data through the Riemannian geometry of positive definite matrices in Python
A Django app that lets you filter data by date range and numeric range in the admin UI
Display, extract, and manipulate PSP firmware inside UEFI images
Generate custom gridfinity bins with AI, from photos of your tools
Traffic scenario definition and execution engine
My template for new Python libraries.
Sublime plugin to prefix your CSS
Project Mu Documentation
A code generator for array-based code on CPUs and GPUs
FAUCET is an OpenFlow controller for multi table OpenFlow 1.3 switches, that implements layer 2 switching, VLANs, ACLs, and layer 3 IPv4 and IPv6 routing.
ORB forcefield models from Orbital Materials
The definitive OpenAI, Claude, MCP, Harness, Evals, and Production Agent Systems learning roadmap.
Stablecoin powered by LLAMMAs
Emulator for Software-Defined Wireless Networks
Offline timezone lookup for WGS84 coordinates, with no polygon simplification - so the answer stays correct at timezone borders.
VyOS command definitions, scripts, and utilities
Asisten crypto berbahasa Indonesia: RAG pengetahuan 267 topik + data pasar realtime (6 bursa, WebSocket, derivatif, on-chain, TVL, DeFi) + tool-calling agent + LLM synthesis
β
[ECCV 2026] FixAnything: 3D-Consistent Rendering Refinement via Video Generative Priors
Performance-tuned llama.cpp for AMD Strix Halo (gfx1151): FA + MoE-prefill fixes with a bundled current Mesa driver. Vulkan and HIP; portable dir, Docker, and distrobox.
β
β
Open recipes, engine patches, and benchmark harnesses for LLM inference on Intel Arc Pro B60/B70 (Battlemage, Xe2). MoE 35B at 160 t/s decode / 7.5K t/s prefill single-stream, 27B at 50~ t/s decode / 1.7K t/s prefill single stream. vLLM XPU MTP unlocked. Muse Glimmer recipe added!!
An open training and inference stack for Jev-style decision models. Train models to score dynamic candidate branches from a shared prefix, with support for high-cardinality choice, calibration, and fast batched inference.
Chat with your documents to create and refine editable architecture diagrams. Local workspace, Excalidraw canvas, optional Tavily search.
Experimental test to freeze video latents while continuing with audio processing, to increase audio quality in low-step turbo MiniMax H3 generations
Polymarket trading bot for prediction markets β browse CLOB markets, watch the order book in the terminal, run edge detection, liquidity provision and cross-market arbitrage strategies with paper trading and risk limits. Educational open-source toolkit β not financial advice. Unofficial community project, not affiliated with Polymarket.
Heterogeneous prefill/decode for DeepSeek-V4-Flash: CUDA prefill (DGX Spark, vLLM) -> Metal decode (Mac Studio, oMLX) over plain 10GbE
DeepSeek-V4.1-Flash EXL3 on one 24 GB RTX 3090 + DDR4 + NVMe: staged-DMA prefill, AVX2 CPU expert tier, elastic VRAM expert cache, Engram on disk, OpenAI API
Community registry of LLM serving-path traps that produce confidently wrong measurements: templates, tool parsers, reasoning fields, quant kernel paths, CUDA toolchains, KV allocation, eval harnesses, versioning. Symptom-first, with the check that catches each.
Qwen3.8-Flash-Next on 2Γ RTX 3090: with 128 GB RAM up to 4,191 tok/s prefill Β· 111.5 tok/s decode (131K prompt), full 256K window at 2,865 tok/s; with 64 GB RAM 3,410 tok/s prefill Β· 84 tok/s decode. New: opt-in uncensored mode (runtime abliteration, no new weights).
DeepSeek-V4-Flash-0731 on 4x CMP 170HX (sm_80): 98 tok/s decode, ~5300 tok/s prefill. DSpark speculative decoding under pipeline parallelism, which vLLM does not support upstream.
LLM Agent for WWTP biotreatment on redefined ASM code library
A Mini-vLLM optimization project for Qwen3-32B TP=2 serving, Prefix Cache, and Triton PagedAttention.
A local app that rents a GPU in your own Colab account and serves Qwen3.8 on it: 27B on an A100-40G, or Flash-Next (125B-A6B MoE) on an A100-80G. The cost is shown before it starts; chat with images; Codex or any OpenAI client uses a fixed local /v1. Auto-stop, CU ledger, live prefill/decode. Windows, macOS, Linux.
Independently analyze ROI, win rate, and PnL using the public Hyperliquid API, detect martingale/grid recovery strategies, and assess maximum drawdown and risk-adjusted returns. Read-only terminal checker β no trading, no API keys. Unofficial community project, not affiliated with Hyperliquid.