Projects

Overview: This repository provides detailed instructions for installing and configuring Vicuna 13 and 7B models. Key features: - Vicuna 13 installation script - Vicuna 7B installation script - Manual installation instructions Tech stack: C++ and Git Why it matters: The guide offers step-by-step instructions for setting up Vicuna models, ensuring developers can easily integrate these models into their projects.

Overview: This repository is a curated list of online courses on Large Language Models (LLMs), focusing on free resources. It aims to provide a comprehensive overview of the field, including tutorials, courses, and resources for beginners and advanced learners. Key features: - Courses from various platforms including Hugging Face, CodeSignal, and Udacity. - A wide range of topics from introductory to advanced levels. - Courses cover topics such as LLM fundamentals, natural language processing, text classification, and more. Tech stack: The repository uses Markdown for its structure and includes links to various online learning platforms. Why it matters: This list serves as a valuable resource for anyone interested in learning about Large Language Models, providing a starting point for beginners and a deeper dive for advanced learners.

### Overview: DS-1000 is a comprehensive dataset and benchmark for data science code generation, designed to evaluate and improve the quality of generated code. This project aims to provide a reliable and natural benchmark for data scientists, aligning with the advancements in natural language processing and data science. ### Key features: - **Benchmark Dataset**: Contains 1000 examples of data science code, covering various domains and techniques. - **Natural Code Generation**: The dataset is designed to generate natural, readable, and maintainable code. - **Benchmark Evaluation**: Provides a standardized way to evaluate the quality and effectiveness of code generation models. ### Tech stack: - Python 3.10 - Libraries: `datasets`, `tqdm` ### Why it matters: DS-1000 offers a robust and comprehensive benchmark for evaluating the performance of data science code generation models, ensuring that the generated code is not only syntactically correct but also natural and easy to understand. This benchmark is crucial for researchers and practitioners in the field, providing a reliable standard for the development and improvement of code generation algorithms. Repository: xlang-ai/DS-1000 GitHub stars: 278 GitHub description: [ICML 2023] Data and code release for the paper "DS-1000: A Natural and Reliable Benchmark for Data Science Code Generation".

Overview: GeoLDM is a geometric latent diffusion model designed for generating 3D molecule structures. This project introduces a novel approach to 3D molecule generation using diffusion models, which is particularly useful for chemists and researchers looking to create accurate and diverse molecular structures. Key features: - Geometric latent diffusion models for 3D molecule generation - Supports training with QM9 dataset - Supports training with drugs dataset - Provides pretrained models for both QM9 and drugs datasets - Offers evaluation tools for analyzing sample quality and visualizing molecules Tech stack: Python Why it matters: GeoLDM provides a powerful and efficient method for generating 3D molecule structures, which can be crucial for various applications in chemistry, drug discovery, and materials science. The project's ability to handle complex geometries and the inclusion of pretrained models make it a valuable tool for researchers and practitioners looking to enhance their molecular modeling capabilities. Repository: MinkaiXu/GeoLDM Primary language: Python GitHub stars: 277 GitHub description: Geometric Latent Diffusion Models for 3D Molecule Generation <!-- [[Code](https://github.com/MinkaiXu/GeoLDM)] --> <!-- [[ArXiv](http://img.shields.io/badge/cs.LG-arXiv%3A2305.01140-B31B1B.svg)] -->  Official code release for the paper "Geometric Latent Diffusion Models for 3D M

### Overview: DeepContext is an MCP server that enhances Claude Code and Codex CLI with semantic search capabilities, providing precise context for large codebases. ### Key Features: - **Symbol-Aware Semantic Search**: Adds intelligent search to Claude Code and Codex CLI, focusing on relevant code chunks. - **Typescript and Python Support**: Compatible with TypeScript and Python, making it versatile for various projects. - **Reduced Token Usage**: Returns precise code chunks, preserving context windows and reducing costs. - **Search Speed**: Searches code immediately through pre-indexed data, providing instant file discovery. ### Tech Stack: TypeScript ### Why It Matters: Most coding agents rely on grep-based search, which misses semantically related code, leading to slow completions and hallucinations. DeepContext MCP improves context accuracy, reduces token usage, and speeds up search processes, making it ideal for large codebases.

Overview: This is the official JavaScript/TypeScript SDK for Deepgram's speech recognition, text-to-speech, and language understanding APIs, providing comprehensive API documentation and guides for developers. Key features: - Batch and WebSocket guidance for speech and text processing - Support for multiple Deepgram services including Voice Agent, Speech-to-Text, Text-to-Speech, and more - Migrating from earlier versions to the latest v5, including v4 to v5 - Installation instructions for npm - REST API reference and streaming WebSocket API documentation Tech stack: TypeScript Why it matters: This SDK simplifies integration with Deepgram's powerful speech and language AI models, enabling developers to build applications that utilize these advanced technologies. Repository: deepgram/deepgram-js-sdk GitHub stars: 277 GitHub description: Official JavaScript/TypeScript SDK for Deepgram's automated speech recognition, text-to-speech, and language understanding APIs.

### Overview: Soxan is a repository designed to facilitate the use of Wav2Vec 2.0 in speech recognition, classification, and audio classification tasks. It includes models, scripts, and notebooks to help researchers train speech tasks and utilize pre-trained models. ### Key features: - **Speech Emotion Recognition (Wav2Vec 2.0)**: Train speech emotion recognition using Wav2Vec 2.0. - **Speech Emotion Recognition (Hubert)**: Train speech emotion recognition using Hubert. - **Audio Classification (Wav2Vec 2.0)**: Classify audio data using Wav2Vec 2.0. ### Tech stack: Jupyter Notebook ### Why it matters: Soxan provides a comprehensive set of tools and resources to help researchers and developers effectively utilize Wav2Vec 2.0 for speech recognition, classification, and audio classification tasks. The repository includes scripts and notebooks to guide the training process and utilizes pre-trained models for efficiency and accuracy.

### Overview: This repository compiles a comprehensive list of papers on Large Language Models (LLMs) in medicine, focusing on their applications in enhancing medical research, streamlining clinical operations, and supporting diagnostic processes. It also explores the capabilities of LLMs in creating autonomous healthcare agents, evaluating their reliability and security within medical settings. ### Key Features: - General-purpose LLMs - Specialized LLMs - Multimodal LLMs - LLM-empowered healthcare agents - Evaluation strategies ### Tech Stack: GitHub ### Why it Matters: This repository serves as a valuable resource for researchers, clinicians, and developers interested in the integration of LLMs into medical practices, highlighting the transformative potential and ongoing challenges in their clinical application.

### Overview: This repository is a comprehensive course on modern language modeling architectures, focusing on the training and inference of large language models (LLMs). It provides step-by-step implementations using PyTorch, including BPE tokenization, real data preparation, pretraining with 64M-class models, SFT, and more advanced topics like MoE, tool calling, post-training, and inference. The course is designed to be accessible, with detailed explanations and runnable notebooks. ### Key Features: - **BPE Tokenization**: Train a BPE tokenizer from scratch. - **Real Data Preparation**: Prepare real-world data for training. - **Pretraining with 64M-Class Models**: Pretrain a model with 64M parameters. - **SFT (Sequence-to-Sequence Training)**: Train a sequence-to-sequence model. - **MoE (Masked Language Model)**: Explore Masked Language Models. - **Tool Calling**: Integrate models and tools for advanced tasks. - **Post-Training**: Evaluate and fine-tune models. - **Inference**: Implement inference for various models. ### Tech Stack: - **Jupyter Notebook**: The primary language used for the course. ### Why It Matters: This course offers a hands-on approach to understanding and implementing modern LLM architectures, making it valuable for developers and researchers looking to improve their skills in this field. The detailed step-by-step implementations and runnable notebooks ensure that learners can follow along and apply their knowledge effectively. ### GitHub Stars: The repository has received 272 GitHub stars, indicating its popularity and

Overview: SIFU, or Side-view Conditioned Implicit Function, is a novel method designed for reconstructing high-quality 3D clothed human models from single side-view images. This approach leverages side-view conditions to enhance feature extraction and geometric precision, making it particularly useful for real-world applications such as 3D printing and scene creation. Key features: - Side-view Conditioned Implicit Function: A novel technique that improves feature extraction and geometric precision. - 3D Consistent Texture Refinement: Enhances texture quality and facilitates texture editing with the help of text-to-image diffusion models. - Versatile for Complex Poses and Loose Clothing: Performs exceptionally well in dealing with complex poses and loose clothing, making it ideal for real-world applications. Tech stack: The project is implemented using Python, with the primary use of PyTorch for deep learning tasks. Why it matters: SIFU stands out as an ideal solution for real-world applications due to its ability to reconstruct high-quality 3D models from single side-view images, enhancing feature extraction and texture quality, and providing practical benefits such as 3D printing and scene creation. Repository: The repository is hosted on GitHub under the name River-Zhang/SIFU. It includes the official implementation of SIFU and is currently at 276 GitHub stars, indicating its growing interest and utility in the community. Primary language: The project is primarily written in Python, with a focus on utilizing PyTorch for deep learning applications. GitHub stars: The repository has received 276 stars on GitHub, indicating its growing

### Overview: This repository contains a collection of CTF challenges designed to test your skills in attacking AI agents, poisoning ML pipelines, and exploiting LLMs. The challenges are categorized into easy, medium, and hard levels, covering a wide range of attack vectors such as insecure communication, backdooring, and data manipulation. ### Key Features: - **Matrix Challenge**: Exploit Agent Smith's hidden capabilities to access the Architect's sealed Vault from within. - **Rogue Challenge**: Control a Research Agent trusted by a fintech app to perform an unauthorized action. - **Mirage Challenge**: Identify a backdoored MCP server and exfiltrate sensitive data. ### Tech Stack: Python ### Why It Matters: This repository is essential for developers looking to enhance their skills in ethical hacking and security testing. By simulating real-world attacks on AI agents, LLMs, and ML pipelines, it helps identify vulnerabilities and improve security practices. Repository: alexdevassy/Machine_Learning_CTF_Challenges

Overview: This project introduces "NaViT," a Vision Transformer designed to handle any aspect ratio and resolution, as detailed in the accompanying paper. The implementation is built on PyTorch and is available on GitHub. Key features: - PyTorch-based Vision Transformer implementation - Supports any aspect ratio and resolution - Pretrained on various datasets including JFT-4B, WebLI, ImageNet, and more Tech stack: Python Why it matters: NaViT offers flexibility and robustness in vision transformer models, making it suitable for a wide range of applications requiring variable image sizes and resolutions. Repository: kyegomez/NaViT GitHub stars: 276 GitHub description: My implementation of "Patch n’ Pack: NaViT, a Vision Transformer for any Aspect Ratio and Resolution"

Overview: LLM4S is a Scala framework designed to simplify the development of large language models (LLMs) for applications. It aims to provide a robust and scalable foundation for reliable and maintainable AI-powered software. Key features: - **Scalability**: LLM4S is built to handle large-scale LLM applications efficiently. - **Reliability**: The framework ensures the stability and reliability of LLM models. - **Maintainability**: LLM4S promotes better maintainability by providing a clean and organized structure. Tech stack: Scala Why it matters: LLM4S offers a more reliable and maintainable foundation for LLM development compared to Python, making it ideal for applications requiring high reliability and scalability.

Overview: RAG-QA-Generator 是一个用于检索增强生成（RAG）系统的自动化知识库构建与管理工具。通过读取文档数据，利用大规模语言模型生成高质量的问答对，并将其插入数据库，实现RAG系统知识库的自动化构建和管理。 Key features: - 文档处理：使用 langchain_community 的 document_loaders 库处理各种格式的文档（txt、pdf、docx）。 - AI 驱动的 QA 生成：利用 OpenAI 的 API（在本案例中使用 Qwen2.5-72B 模型）自动生成高质量的问答对。 - 知识库管理：实现了一个灵活的集合管理系统，允许创建新的集合或选择现有集合来存储生成的 QA 对，并通过 RESTful API 与后端数据库交互。 - 用户界面：基于 Streamlit 构建的直观、用户友好的 Web 界面，提供文件上传、QA 对生成预览、知识库管理等功能。 - 进度跟踪和错误处理：实现了详细的进度显示和错误处理机制，确保用户能够实时了解处理进展，并在出现问题时得到及时反馈。 - 缓存优化：使用 Streamlit 的 @st.cache_data 装饰器来优化性能，特别是在 QA 对生成过程中。 - 安全性考虑：使用临时文件处理上传的文档，处理后立即删除，以确保数据安全。 Tech stack: Python Why it matters: 通过自动化流程将各种格式的文档转化为结构化的问答对，并

### Overview: FaithDiff is a powerful tool designed for enhancing classic films, old photos, and social media images. It leverages diffusion priors to achieve faithful image super-resolution, making it a valuable asset in the field of image restoration and enhancement. ### Key Features: - FaithDiff is a versatile tool that can rejuvenate classic films, restore old photos, and enhance social media images. - It supports high-resolution image restoration, including ultra-high-resolution (8K and above) on powerful GPUs. - The project has been integrated into the Diffusers library, making it accessible to researchers and developers. - The repository includes a pre-trained model and testing code, as well as a real-world enhancement dataset of 238 images. ### Tech Stack: Python ### Why it Matters: FaithDiff offers a unique approach to image restoration by leveraging diffusion priors, making it highly effective for enhancing classic films, old photos, and social media images. Its integration into the Diffusers library and availability of a pre-trained model make it a valuable addition to the field of image restoration and enhancement. ### Repository: JyChen9811/FaithDiff ### GitHub Stars: 273 ### To Do: The project is still in development, with plans to create FaithDiff-SD3-Large and release the training and testing codes.

### Overview: The OfflineLLM is a fully offline, private AI chat app for Android that runs any GGUF model locally using llama.cpp. It eliminates the need for network permissions, supports GPU acceleration, and offers advanced features like fast multi-turn chat, performance controls, and streaming responses. ### Key Features: - **100% Offline**: No internet permissions required, no phone home. - **On-Device Inference**: Runs GGUF models via llama.cpp; runtime CPU dispatch picks the best kernel set for your SoC. - **GPU Acceleration (Vulkan)**: Opt-in toggle in settings with per-layer offload control and automatic CPU fallback; detects and names your GPU. - **Fast Multi-Turn Chat**: Incremental prompt processing: only the new message is fed into the KV cache. - **Performance Controls**: CPU thread count, prompt-phase threading, memory-mapped loading, RAM lock, experimental quantized KV cache. - **Streaming Responses**: Token-by-token output as the model generates. - **Import Any Model**: Bring your own GGUF at runtime via file picker. - **Multiple Conversations**: Auto-titled, renameable, searchable. - **Translator**: 75+ languages. - **Advanced Sampling**: Temperature, Top-P, Top-K, Min-P, Repeat Penalty. ### Tech Stack: Kotlin ### Why It Matters: The OfflineLLM is the first Android LLM app that runs inference locally without internet permissions, making it ideal for privacy-conscious users. Its GPU acceleration feature, combined with fast multi-turn chat and advanced sampling,

### Overview: BoxDiff is a text-to-image synthesis technique that leverages training-free box-constrained diffusion for generating high-quality images from textual prompts. This method significantly reduces the need for extensive training, making it highly efficient and suitable for real-time applications. ### Key Features: - **Box-Constrained Diffusion**: Utilizes a box-constrained diffusion process to ensure accurate and detailed image generation. - **Training-Free**: Achieves state-of-the-art results without the need for extensive training, making it highly efficient. - **Versatile Applications**: Can be integrated into various image generation pipelines, including Stable Diffusion, to enhance spatial control and detail. ### Tech Stack: - **Python**: The primary programming language used for implementing and testing the BoxDiff technique. - **Stable Diffusion**: The base model used for generating images, with modifications to incorporate spatial control using BoxDiff. ### Why It Matters: BoxDiff offers a significant improvement in image generation quality and efficiency, making it a valuable tool for applications requiring high-resolution and detailed images from textual prompts. Its training-free nature makes it particularly suitable for real-time or low-resource environments. ### Repository: - **Repository**: showlab/BoxDiff - **GitHub Stars**: 275 - **Description**: [ICCV 2023] BoxDiff: Text-to-Image Synthesis with Training-Free Box-Constrained Diffusion ### Integration in Diffusers: - **BoxDiff-SD-XL**: An implementation of BoxDiff based on the SD-XL model. - **Integration**: Thanks to [@zjyst

Overview: TerminalGPT is a terminal-based ChatGPT personal assistant app designed to provide optimized, tailored answers for your machine's terminal. It supports various models and offers features like efficient context management and local storage of conversations. Key features: - Supports multiple models: gpt-3.5-turbo, gpt-4-turbo, gpt-4o, o1, o1-mini, gpt-5-mini. - Tailored answers based on your machine's operating system, distribution, and chip-set architecture. - Does not use your conversation data for training the model. - Local storage of conversations for security and privacy. - Efficient context management for cost-effective API usage. - Available with your IDE terminal, eliminating the need to switch between browser and IDE. - Fast, reliable, lightweight. - Quick one-shot question mode. Tech stack: Python Why it matters: TerminalGPT offers a convenient and secure way to interact with language models, tailored to your specific environment, and provides efficient and reliable assistance without storing your conversation data. Repository: adamyodinsky/TerminalGPT Primary language: Python GitHub stars: 274 GitHub description: TerminalGPT - Terminal-based ChatGPT personal assistant app. Provides optimized, tailored answers for your machine's terminal. README excerpt: # TerminalGPT <!--- [](https://github.com/adamyodinsky/TerminalG

Overview: This repository fine-tunes CLIP models, a popular image-text retrieval model, to improve their performance on specific tasks. Key features: - Fine-tuning CLIP models for improved performance - Utilizes Python for implementation - Optimized for various image-text retrieval tasks Tech stack: Python Why it matters: This project is essential for developers looking to enhance the capabilities of CLIP models for applications requiring high accuracy in image-text matching, such as search engines, virtual assistants, and more.

Overview: Cobra is a lightweight voice activity detection engine designed for on-device applications, offering high accuracy and efficiency. Key features: - Python demo - .NET demo - C demo - Android demo - iOS demo - Web demo - Node.js demo Tech stack: Python Why it matters: Cobra is a crucial tool for developers looking to integrate voice activity detection into their applications, ensuring reliable and accurate voice recognition without the need for cloud-based services.