Projects

### Overview: PromptSource is a toolkit designed to create, share, and use natural language prompts. It provides a growing collection of prompts, including over 2,000 English prompts for 170+ datasets, and a simple API for easy access and use. ### Key Features: - **Prompt Templates**: Prompts are functions that map examples from datasets to natural language inputs and target outputs. - **Standalone Files**: Prompts are saved in structured files written in Jinja, a simple templating language. - **API**: A simple API for creating, sharing, and using prompts. - **Datasets**: Supports a variety of datasets, including Hugging Face Datasets. ### Tech Stack: - Python ### Why It Matters: The toolkit is essential for researchers and engineers looking to leverage large language models for zero-shot generalization, as demonstrated by GPT-3, FLAN, and T0. Prompts are crucial for enabling these models to perform well on new tasks without extensive training, making PromptSource a valuable tool in the NLP community. ### Repository: - **Repository**: bigscience-workshop/promptsource - **Stars**: 3041 - **Description**: Toolkit for creating, sharing, and using natural language prompts. ### Setup: - **Install Locally**: Use `pip install -e .` to install the package locally. - **Stability**: Requires Python 3.7 environment for stability reasons, but can be removed in the `setup.py` file for local installation. ### How to Use Prompts: -

Overview: ChainForge is an open-source visual programming environment designed for battle-testing prompts to Large Language Models (LLMs). It enables rapid-fire, quick-and-dirty comparison of prompts, models, and response quality, providing insights beyond ad-hoc chatting with individual LLMs. Key features: - Query multiple LLMs at once to test prompt ideas and variations quickly and effectively. - Compare response quality across prompt permutations, across models, and across model settings to choose the best prompt and model for your use case. - Setup evaluation metrics (scoring function) and immediately visualize results across prompts, prompt parameters, models, and model settings. - Use AI to streamline this entire process, including creating synthetic tables and input examples with built-in genAI features, or supercharging writing evals by prompting a model to give you starter code. Tech stack: ChainForge is built on ReactFlow and Flask. Why it matters: ChainForge provides a powerful tool for evaluating and refining prompts for LLMs, allowing developers to quickly test and optimize their interactions with these models.

Overview: Linly-ChatFlow 是一个基于 LLaMA 和 Falcon 的中文对话模型，提供多种语言模型版本，包括 Linly-Chinese-LLaMA-2 和 Linly-OpenLLaMA。 Key features: - Linly-Chinese-LLaMA-2 (70B) - Linly-OpenLLaMA (3B、7B、13B) - Linly-Chinese-LLaMA-1 (3B、7B、13B) - Linly-Chinese-LLaMA-2 (7B、13B) Tech stack: Python Why it matters: Linly-ChatFlow 是一个强大的中文对话模型，适用于多种应用场景，如聊天机器人、客服系统等。 Repository: CVI-SZU/Linly GitHub stars: 3035 GitHub description: 提供中文对话模型 Linly-ChatFlow、中文基础模型 Chinese-LLaMA (1-2)、Chinese-Falcon 及其训练数据。

Overview: DreamCraft3D is an advanced hierarchical 3D content generation method that leverages a 2D reference image to produce high-fidelity and coherent 3D objects. It addresses the consistency issue in existing works by using a 3D prior and personalized diffusion models. Key features: - Hierarchical 3D generation - 2D reference image guidance - Score distillation sampling - Bootstrapped Diffusion Prior - Personalized diffusion model training Tech stack: Python Why it matters: DreamCraft3D advances the state-of-the-art in 3D content generation by producing photorealistic renderings and ensuring coherence in 3D objects, making it a valuable tool for applications in computer graphics, architecture, and more. GitHub stars: 3033

Overview: This project is a community tool that keeps track of registration deadlines for Machine Learning Summer Schools worldwide. Key features: - Central repository for summer school deadlines - Upcoming and happening schools - Detailed information on past schools in the web archive Tech stack: HTML Why it matters: It serves as a valuable resource for students and professionals interested in machine learning summer schools, helping them plan their studies effectively.

### Overview: Pentest Swarm AI is an open-source, autonomous penetration testing tool that uses a swarm of AI agents to automate and enhance the process of identifying vulnerabilities and exploiting them. It supports bug bounty, continuous monitoring, and CTF modes, leveraging the power of multiple agents working together in parallel to achieve machine-speed efficiency. ### Key Features: - **Swarm of AI Agents:** Hundreds of agents working in parallel to scan, classify, exploit, and report on vulnerabilities. - **Real Swarm Model:** Agents communicate through a shared stigmergic blackboard, ensuring that each agent can quickly respond to relevant findings. - **Versatile Mode Support:** Supports bug bounty, continuous monitoring, and CTF modes, making it suitable for various security roles. - **Scoring and Reporting:** Adapts attack-path scoring to assess and report on vulnerabilities effectively. - **Built with Go:** Utilizes the Go programming language and integrates with 7 native security tools. ### Tech Stack: - **Go:** The primary language used for development. - **7 Native Security Tools:** Integrated with various security tools to enhance functionality. ### Why It Matters: - **Open Source and Self-Hosted:** Supports pentesters, bug bounty hunters, red teamers, and security researchers, providing a flexible and customizable solution. - **Machine Speed Efficiency:** Hundreds of agents working in parallel to achieve high-efficiency scanning and exploitation. - **Versatile and Scalable:** Supports various modes of operation, making it suitable for different security needs and environments. ### Repository: - **Armur-Ai/P

Overview: Stats is a comprehensive Golang statistics library package that provides a wide range of statistical functions without any dependencies. Key features: - Median, Round, and other basic statistical calculations - Loading raw data from various sources - Functions for calculating correlation, covariance, and more - Descriptive statistics including mean, variance, and more - Clipping and transforming data - Calculating cumulative sums, products, and max/min values - Distance metrics like Euclidean and Chebyshev - Time series analysis with EWMA and moving averages Tech stack: Go Why it matters: Stats simplifies statistical computations for developers by providing a robust set of functions without the need for external dependencies, making it ideal for projects requiring statistical analysis without the overhead of additional libraries.

### Overview: Markstream is a multi-framework streaming Markdown renderer for AI chat, LLM token streams, and more. It supports Vue, React, Svelte, Angular, and more, with features like Mermaid, KaTeX, and safe HTML rendering. ### Key features: - **Vue/Nuxt, React/Next.js, Svelte, Angular, and Vue 2 compatibility**. - **Stable partial Markdown states** for Vue/Nuxt/VitePress. - **Mobile WebView rendering** and safe component rendering. - **Progressive heavy blocks** for enhanced user experience. - **Safe HTML** and low-jitter updates. ### Tech stack: Vue 3 / Nuxt / VitePress. ### Why it matters: Markstream is essential for applications that need to render Markdown in real-time, such as AI chat interfaces, where the Markdown content is continuously updated and the user experience should be smooth and stable.

### Overview: This repository is a curated list of recent papers and resources related to deep learning and deep reinforcement learning. The papers are sorted by year to provide a chronological view of the latest developments in the field. The repository includes a table of contents for easy navigation, with sections dedicated to papers, model zoo, pretrained models, courses, books, tutorials, software, applications, and a corpus. ### Key Features: - Papers: A comprehensive list of recent papers on deep learning and deep reinforcement learning. - Model Zoo: A collection of pre-trained models for various tasks. - Pre-Trained Model: A repository of pretrained models and embeddings. - Courses: A list of courses and tutorials on deep learning and reinforcement learning. - Books: A selection of books on deep learning and reinforcement learning. - Tutorials: A collection of tutorials and guides for deep learning and reinforcement learning. - Software: A repository of software tools and libraries used in deep learning and reinforcement learning. - Applications: A list of real-world applications of deep learning and reinforcement learning. - Corpus: A collection of resources and papers related to deep learning and reinforcement learning. ### Tech Stack: The repository is primarily written in Markdown and hosted on GitHub. ### Why It Matters: The papers and resources listed in this repository are considered important or popular because they represent the cutting-edge research in deep learning and reinforcement learning. The model zoo and pretrained models provide a valuable resource for researchers and practitioners looking to quickly get started with these technologies. The courses and tutorials offer educational resources that can help anyone from beginners to advanced learners to deepen their understanding of these fields

Overview: Deepnote is a data notebook designed for the AI era, offering a sleek UI, new blocks, and seamless integration with various data sources. It's used by 500,000+ professionals and supports Python, R, and SQL. Key features: - **Sleek UI**: A modern and intuitive design. - **New Blocks**: Enhances notebook functionality with SQL, inputs, charts, etc. - **Native Data Integrations**: Seamless integration with various data sources. - **AI Agent**: Enhances collaboration and productivity. - **Cloud Support**: Supports real-time collaboration, agent, and deployable data apps. Tech stack: TypeScript Why it matters: Deepnote offers a human-readable format, a block-based architecture, local and cloud support, and seamless conversion between Jupyter and Deepnote notebooks, making it a powerful tool for data professionals. Repository: deepnote/deepnote GitHub stars: 3025 GitHub description: Deepnote is a drop-in replacement for Jupyter with an AI-first design, sleek UI, new blocks, and native data integrations. Use Python, R, and SQL locally in your favorite IDE, then scale to Deepnote cloud for real-time collaboration, Deepnote agent, and deployable data apps.

### Overview: AlphaTree 是一个旨在为AI研究和应用工程师提供全面资源的项目。它涵盖了从数学到计算机科学，再到人工智能的三千年演进历程，包括经典网络结构、论文发展路线图、可视化网络结构、交互式算法动画和论文地铁图。项目通过丰富的资源帮助读者快速跟进前沿技术，提高应用能力。 ### Key Features: - **经典网络结构**：统一画法的深度学习经典网络结构。 - **论文发展路线图**：涵盖从1943到2026年的AI发展里程碑。 - **可视化网络结构**：展示网络结构的可视化图。 - **交互式算法动画**：动画展示算法的执行过程。 - **论文地铁图**：展示论文在不同领域的连接关系。 ### Tech Stack: 纯前端单页，无后端、无需登录。 ### Why it Matters: AlphaTree 是一个全面的资源库，帮助AI研究和应用工程师快速跟进前沿技术，提高应用能力。它通过丰富的资源和互动方式，使读者能够轻松地学习和理解AI的发展历程，从而在面试和项目中脱颖而出。 ### Repository: - **Repository**: weslynn/AlphaTree-graphic-deep-neural-network - **Stars**: 3025 - **Description**: 2019年的静态图解合集，2026年重做成可交互的AI时间线站点，收录了148个里程碑，包含深度学习、GAN、NLP、大数据等领域的

Overview: Fetches source code for npm packages to enhance AI coding agent capabilities. Key features: - Access to any package's source code - Works with npm, PyPI, crates.io, and GitHub - Provides quick start instructions and CLI usage examples Tech stack: Rust Why it matters: Enhances AI coding agents with deeper context and improved development experience. Repository: vercel-labs/opensrc GitHub stars: 3024 GitHub description: Fetch source code for npm packages to give AI coding agents deeper context

### Overview: Super-SloMo is a PyTorch implementation of the "Super SloMo: High Quality Estimation of Multiple Intermediate Frames for Video Interpolation" paper by Jiang et al. It provides a method to interpolate and slow down videos by generating intermediate frames, achieving better quality than traditional techniques. The code is designed for both CPU and GPU environments and includes a pretrained model trained on the Adobe240fps dataset. ### Key features: - **Interpolation and Slow Motion**: Generates intermediate frames to slow down videos. - **Pretrained Model**: Available for download on Google Drive. - **GPU and CPU Compatibility**: Works with both CPU and GPU environments. - **Training Data Preparation**: Uses `ffmpeg` to extract frames from videos. - **Training Script**: Includes a script to create dataset and train the model. - **Evaluation**: Results are provided on the UCF101 dataset using specific metrics. ### Tech stack: Python ### Why it matters: Super-SloMo offers a significant improvement in video quality by generating intermediate frames, making it a valuable tool for slow-motion video applications. Its pretrained model trained on the Adobe240fps dataset provides a starting point for further research and development.

### Overview: This repository is a curated list of resources on Transformers in Time Series, focusing on the latest advancements in using Transformers to model time series data. It serves as a comprehensive overview of the field, including surveys, papers, and code snippets. ### Key Features: - **Transformers in Time Series: A Survey** (IJCAI'23 Survey Track) - **Author**: Qingsong Wen, Tian Zhou, Chaoli Zhang, Weiqi Chen, Ziqing Ma, Junchi Yan, Liang Sun - **Publication**: International Joint Conference on Artificial Intelligence (IJCAI) - **Details**: A survey paper that comprehensively summarizes the recent advances of Transformers for modeling time series data. - **Taxonomy of Transformers for Time Series Modeling** - **Image**:  (Not specified) - **Description**: A taxonomy of Transformers for time series modeling, providing a clear overview of the field. - **Application Domains of Time Series Transformers** - **Forecasting**: - **Time Series Forecasting**: - **CARD**: Channel Aligned Robust Blend Transformer for Time Series Forecasting, ICLR 2024. - **Pathformer**: Multi-scale Transformers with Adaptive Pathways for Time Series Forecasting, ICLR 2024. - **GAFormer**: Enhancing Timeseries Transformers Through Group-Aware Embeddings, ICLR 2024. - **Transformer-Modulated Diff

### Overview: Welcome to our GitHub repository featuring 50+ Python applications with only 10 lines of code! This repository offers a wide range of topics such as Machine Learning, Deep Learning, GUI, Computer Vision, and API development. Each application is designed to be simple and concise, making it easy to understand and modify. Whether you're a beginner or an experienced developer, these applications are perfect for learning and experimenting with Python. ### Key Features: - Voice Recorder - Password Protect PDF - Merge Multiple PDFs - Windows Notification - Audio Visualization Tool - Random Password Generator - Extract mp3 from mp4 - Link Shortener and Extractor - Terminal Tricks - Birthday Reminder - Audiobook - Alarm - Schedule YouTube Video ### Tech Stack: Python ### Why It Matters: These applications are designed to be simple and concise, making them perfect for learning and experimenting with Python. Whether you're a beginner or an experienced developer, these applications are perfect for learning and experimenting with Python.

Overview: Torch-TensorRT is a powerful tool that allows developers to optimize PyTorch models for NVIDIA GPUs, significantly improving inference performance by up to 5x compared to eager execution. Key features: - Easily achieve the best inference performance for any PyTorch model on the NVIDIA platform. - Optimizes models using TensorRT for faster inference. - Supports both PyTorch and C++ deployment. - Compatible with PyTorch 2.14 and CUDA 13.4. - Available on PyPI and the PyTorch package index. - Includes Linux, Windows, and NVIDIA NGC PyTorch Container wheels. Tech stack: PyTorch, TensorRT, CUDA, PyTorch, PyTorch, PyTorch, PyTorch, PyTorch. Why it matters: Developers can significantly boost the performance of their PyTorch models by leveraging the optimized TensorRT backend, making the models run faster and more efficiently on NVIDIA GPUs.

### Overview: This repository contains a comprehensive map of natural language processing (NLP) tasks, providing an overview of various NLP tasks and references. It aims to serve as a starting point for those interested in diving deeper into specific NLP tasks, while also encouraging collaboration to update and expand the repository. ### Key Features: - Anaphora Resolution - Automated Essay Scoring - Automatic Speech Recognition - Automatic Summarization - Coreference Resolution ### Tech Stack: Python and TensorFlow. ### Why It Matters: This repository serves as a valuable resource for anyone interested in NLP, providing a clear overview of various tasks and references. It is a starting point for those looking to explore specific NLP tasks and a call to action for collaboration to keep the repository updated and expanded.

### Overview: This repository serves as a sandbox for researching and training deep learning networks, primarily focusing on computer vision tasks. It contains implementations of various classification, segmentation, detection, and pose estimation models, along with scripts for training, evaluating, and converting models. ### Key Features: - **Models**: Implementations of popular deep learning models such as AlexNet, VGG, ResNet, and others. - **Frameworks**: Ported to multiple frameworks including MXNet/Gluon, PyTorch, Chainer, Keras, and TensorFlow. - **Datasets**: Pretrained weights are loaded automatically for ImageNet-1K, CIFAR-10, CIFAR-100, SVHN, and CUB-200-2011. - **Documentation**: Detailed documentation for each framework, including examples of automatic weight loading. ### Tech Stack: - Python ### Why It Matters: This repository provides a comprehensive platform for researchers and developers to explore and implement deep learning models, making it valuable for both academic and practical applications in computer vision tasks.

Overview: Weld is a high-performance runtime designed specifically for data-intensive applications, optimizing across libraries and functions to improve overall application performance. Key features: - Weld optimizes across multiple libraries and frameworks by expressing core computations in a common intermediate representation. - It optimizes and evaluates workflow computations only when needed, reducing data movement and improving performance. - Supports modern analytics workflows that combine functions from different libraries and frameworks. Tech stack: Rust, LLVM/Clang++. Why it matters: Weld addresses the performance bottleneck in modern data-intensive applications by optimizing workflow computations, leading to significant improvements in application performance and efficiency. Repository: weld-project/weld Primary language: Rust GitHub stars: 3011 GitHub description: High-performance runtime for data analytics applications.

### Overview: This project implements the Simple Baselines for Human Pose Estimation and Tracking method from the ECCV2018 paper, providing baseline methods that are simple and effective, suitable for inspiring and evaluating new ideas in the field. ### Key features: - **Model Architecture**: Supports multiple architectures including ResNet-50, ResNet-101, and ResNet-152. - **Benchmark Results**: Achieves high accuracy on various benchmarks, including MPII and COCO keypoints. - **Flip Test**: Utilizes flip test for improved performance. - **COCO Validation**: Achieves an AP of 74.3 on COCO keypoints valid dataset. - **COCO Validation with Detector**: Achieves an AP of 56.4 on COCO validation dataset. ### Tech stack: Python ### Why it matters: This project provides a simple yet effective baseline method for human pose estimation and tracking, making it a valuable resource for researchers and developers looking to improve their own implementations.