modal-labs/gpu-glossary

GPU documentation for humans

This repository holds the source files for the Modal GPU Glossary, an online dictionary of terms related to GPU programming, with a focus on the NVIDIA GPUs available on the Modal platform.

What's inside: The glossary content lives in the gpu-glossary folder. The rest of the repository contains supporting files (like license texts, build tools, etc.).

Licensing: Two licenses apply. All files inside the gpu-glossary folder are under the Creative Commons Attribution 4.0 International (CC BY 4.0) license. All other files in the repository are under the MIT License. Each license file is included.

Translations: The README links to a community-maintained Chinese (简体中文) translation of the glossary at miter6/gpu-glossary-zh.

Related

  • Project
    SharpAI/SwiftLM

    A native Swift, OpenAI-compatible inference server for MLX models on Apple Silicon, featuring TurboQuant (4× KV-cache compression), MTP speculative decoding, and SSD streaming for large MoE models, with extensive benchmarks on M-series chips.

  • Project
    bstnxbt/dflash-mlx

    A speculative decoding implementation for Apple Silicon that uses block-diffusion draft models to accelerate LLM inference without losing output quality.

  • Project
    tiiuae/Falcon-Perception

    Falcon‑Perception is an open‑source, PyTorch/MLX inference library for the Falcon multimodal transformer that can answer natural‑language queries with object detection, instance segmentation, or OCR. It provides highly‑optimized paged and batch engines, a FastAPI server, a Streamlit UI, Docker/vLLM deployment for OCR, and Colab notebooks. The 270 M‑parameter Falcon‑OCR 1.5 model delivers end‑to‑end OCR quality comparable to much larger VLMs while staying three‑times smaller.

  • Project
    sgasser/pasteguard

    PasteGuard is a self‑hosted proxy that redacts personal data and secrets from requests before they reach LLM providers (ChatGPT, Claude, Gemini, etc.). It runs via Docker, offers a dashboard, and supports both masking‑only and routing‑to‑local‑LLM modes.