The repository provides a comprehensive ecosystem for the MiniCPM-V and MiniCPM-o series of multimodal large language models (MLLMs). It enables efficient on-device inference and real-time full-duplex multimodal interaction (vision + audio + text) with low latency. The core contributions include the OmniLMM architecture with intra-ViT early compression for visual encoding, mixed 4x/16x visual token compression, and support for streaming audio/video inputs. The system includes evaluation frameworks for multimodal benchmarks (MMBench, Video-MME, OmniDocBench), fine-tuning utilities (LoRA and full model), and deployment tools for various platforms (iOS, Android, HarmonyOS, web) and inference backends (vLLM, SGLang, llama.cpp, Ollama). Key entry points are the model inference code in omnilmm/model/, evaluation scripts in eval_mm/vlmevalkit/, fine-tuning in finetune/, and web demos in web_demos/ demonstrating real-time chat with multimodal inputs.
This repository is a collection of web demos and examples for MathJax version 4, the JavaScript math rendering library. It demonstrates various ways to integrate MathJax into web pages, including standard TeX/MathML input with CHTML/SVG output, equation numbering and referencing, interactive reveal/toggle steps, speech generation, and compatibility with MathJax v2 configurations. The repo also contains a substantial custom font extension named 'mathjax-newcm' (New Computer Modern) that provides extensive glyph metrics and SVG path data for mathematical typesetting, with a specific focus on braille output (the 'euro-braille' subdirectory). This font data is split into dynamic chunks loaded on demand by MathJax. The codebase is primarily illustrative—each demo is a self-contained HTML page with accompanying JavaScript. The custom component and TeX extension examples show how to build tailored MathJax bundles using internal APIs. The configuration converter is a standalone utility that parses legacy MathJax v2 configuration objects and generates equivalent v4 configuration. While the demos are functional and well-organized, the font data files are minified/generated artifacts that are difficult to maintain manually, and several parts rely on MathJax internal APIs that may change across versions. The euro-braille and euro-braille-old directories contain near-duplicate font data with version drift, indicating a maintenance challenge. Overall, this repo serves as a practical reference for MathJax v4 integration patterns and custom font development, but it is not a production library itself. The most reusable components are the custom component assembly pattern, the TeX extension template, and the configuration converter utility.
This repository is a collection of standalone web application demos from Intel showcasing the W3C Generic Sensor API (Accelerometer, Gyroscope, Magnetometer, AmbientLightSensor, OrientationSensor). The demos are built with Polymer 2/3, LitElement, Three.js, and A-Frame, and include a sensor polyfill, a sensor tester with automated test cases, a punch speed meter, a VR button using magnetometer, 360-degree panorama and video viewers, and an ambient light map. The code is functional but outdated, relying on deprecated Polymer patterns and old library versions. The primary value lies in the sensor polyfill implementation and the walking detection algorithm, which are self-contained and demonstrate non-obvious engineering skill. The repo was migrated to the W3C organization and is no longer actively maintained.