Projects

Overview: sqz is a context compression and session deduplication tool designed for AI coding agents. It optimizes AI model inputs by compressing tool outputs and deduplicating repeated references, significantly reducing the number of LLM calls and improving overall efficiency. Key features: - Compresses tool output before it enters the model. - Deduplicates repeated references to 13-token refs. - Real-world stats: 3,003 compressions, 178,442 tokens saved, 24.7% average reduction, up to 92% on output already containing relevant information. Tech stack: Rust, zero LLM calls. Why it matters: sqz helps AI coding agents optimize their input by reducing the number of LLM calls, thereby improving performance and efficiency. It is particularly useful for applications that heavily rely on context, such as CI/CD pipelines, where context-heavy tool outputs can significantly impact model performance. Repository: ojuschugh1/sqz Primary language: Rust GitHub stars: 662 GitHub description: Context compression for AI coding agents: compresses tool output before it enters the model, dedups repeats to 13-token refs. Claude Code, Cursor, Codex, Kiro, Zed, any MCP client. Rust, zero LLM calls. README excerpt: <p align="center"> <pre> ███████╗ ██████╗ ███████╗ ██╔════╝██╔═══██╗╚══███╔╝ ███████╗██

Overview: pyLLMSearch is an advanced question-answering system designed for interacting with local documents, powered by a custom Large Language Model (LLM). It offers improved parsing, hybrid search, deep linking, and more, with a focus on enhancing the quality of results. Key features: - Fast parsing and embedding of medium-sized document bases - Incremental updates for new documents - Built-in parsers for `.md`, `.pdf`, and `.docx` formats - Support for various document formats using Unstructured pre-processor - FastAPI-based API and MCP server for communication with RAG - Deep linking into document sections - Interoperability with OpenAI and HuggingFace models - SSE MCP Server for interface with popular MCP clients - Hybrid search and re-ranking with Dense and Sparse embeddings - Support for table parsing and image parsing - Optional HyDE for improved results on certain topics Tech stack: Python, Jupyter Notebook Why it matters: pyLLMSearch provides a robust and efficient RAG system with advanced features, making it suitable for developers looking to enhance their document querying capabilities. Repository: snexus/llm-search GitHub stars: 663

### Overview: Energy-Based Transformers (EBTs) are a novel approach that enables generalizable reasoning and scalable learning across any problem or modality. This project showcases the scalability of EBTs by surpassing feed-forward Transformers in terms of performance across various dimensions, including data, depth, parameters, FLOPs, and more. EBTs demonstrate superior generalization capabilities compared to existing models. ### Key Features: - **Scalability**: Becoming the first approach to outscale feed-forward Transformers across modalities and axes including data, depth, parameters, FLOPs, etc. - **Generalizable Reasoning**: Enables generalizable reasoning and System 2 Thinking on any problem or modality. - **Minimal Training Loop**: Demonstrates the effectiveness of EBTs through a simple training loop example. ### Tech Stack: - Python ### Why it Matters: EBTs represent a significant advancement in the field of artificial intelligence, offering scalable and generalizable solutions for a wide range of applications. Their ability to outscale existing models and achieve superior performance across various metrics makes them a promising tool for researchers and practitioners alike. ### Repository: Repository: alexiglad/EBT GitHub Stars: 662 GitHub Description: PyTorch Code for Energy-Based Transformers paper -- generalizable reasoning and scalable learning ### Setup: To set up the environment, use the following commands: ```bash conda create -n ebt python=3.12 conda activate ebt pip install -r requirements.txt ``` For more detailed requirements, use: ```bash

Overview: OneDiffusion is an official implementation of the OneDiffusion paper presented at CVPR 2025, a versatile large-scale diffusion model designed for bidirectional image synthesis and understanding across various tasks. Key features: - Bidirectional image synthesis - Support for diverse tasks - Supports text-to-image, image-to-image, and image-to-text conversions - Provides a prompt guide and supported tasks Tech stack: Python Why it matters: OneDiffusion offers a powerful and versatile tool for generating high-quality images and text descriptions, making it valuable for applications in art, design, and creative industries. Repository: lehduong/OneDiffusion GitHub stars: 664

Overview: Open-dLLM is an open-source project that provides a comprehensive stack for diffusion language models, including pretraining, evaluation, inference, and checkpoints. This project is designed to be transparent and reproducible, making it accessible to developers and researchers alike. Key features: - Pretraining pipeline - Inference scripts - Evaluation suite - Weights and checkpoints on Hugging Face Tech stack: Python Why it matters: Open-dLLM offers a complete set of tools for diffusion language models, enabling developers to work seamlessly from raw data to checkpoints and beyond, ensuring reproducibility and transparency in their research and development processes.

### Overview: pysentimiento is a Python toolkit designed for Sentiment Analysis and Social Natural Language Processing (NLP) tasks. It leverages Transformer-based models to provide accurate and efficient analysis across various languages. ### Key Features: - **Sentiment Analysis**: Supports English, Spanish, Italian, and Portuguese. - **Hate Speech Detection**: Detects hate speech in multiple languages. - **Irony Detection**: Identifies irony in various languages. - **Emotion Analysis**: Analyzes emotions in English. - **NER & POS Tagging**: Performs Named Entity Recognition and Part-of-Speech tagging. - **Contextualized Hate Speech Detection**: Detects hate speech with contextual understanding. - **Targeted Sentiment Analysis**: Analyzes sentiment with targeted detection. ### Tech Stack: The toolkit is built using Jupyter Notebook and relies on Python libraries for sentiment analysis and NLP tasks. ### Why It Matters: pysentimiento simplifies the process of sentiment analysis and NLP tasks, making it accessible for developers and researchers. Its ability to handle multiple languages and contextualized hate speech detection ensures that it remains a valuable tool in the field of social media analysis and sentiment analysis. ### Repository: The repository for pysentimiento is hosted on GitHub, with 662 GitHub stars, indicating its popularity and usefulness in the community. ### Getting Started: To get started with pysentimiento, simply install it via `pip install pysentimiento` and begin using it in your projects. The provided examples in the notebooks demonstrate how to perform sentiment analysis, hate speech detection, and emotion analysis

Overview: ReservoirPy is a simple and flexible library designed for implementing Reservoir Computing architectures, specifically Echo State Networks (ESNs). It provides a straightforward way to create complex architectures with multiple reservoirs, feedback loops, offline and online training, parallelization across sequences, and advanced learning rules. Key features: - Easy creation of complex architectures with multiple reservoirs, including deep reservoirs and multiple readouts. - Support for feedback loops. - Offline and online training options. - Parallelization across sequences. - Sparse matrix computations for efficient computation. - Advanced learning rules like Intrinsic Plasticity, Local Plasticity, and NVAR. - Interfacing with scikit-learn models. - Graphical tools for easily exploring hyperparameters using the hyperopt library. Tech stack: Python Why it matters: ReservoirPy simplifies the implementation of Reservoir Computing architectures, making it accessible for researchers and developers alike. Its flexibility and advanced features make it suitable for a wide range of applications, from time series prediction to sequence classification. Repository: reservoirpy/reservoirpy GitHub stars: 662 GitHub description: A simple and flexible code for Reservoir Computing architectures like Echo State Networks.

### Overview: MICA is a project that aims to reconstruct human faces using metrical reconstruction techniques, published at the 2022 European Conference on Computer Vision (ECCV). The repository contains a face tracker and a dataset for training and testing the model. ### Key features: - **Face Tracker**: A tool for tracking faces in images. - **Dataset**: A collection of 2300 subjects with a common FLAME topology, used for training and testing the model. - **Training Scripts**: Scripts to generate Arcface input images and the complete list of training images. ### Tech stack: Python ### Why it matters: MICA provides a comprehensive solution for face reconstruction, which is crucial in various applications such as augmented reality, face recognition, and more. The metrical reconstruction technique ensures accurate and reliable face reconstruction, making it valuable for research and development in computer vision and related fields.

### Overview: mLLMCelltype is a multi-LLM consensus framework designed to automate cell type annotation in single-cell RNA sequencing (scRNA-seq) data. By integrating multiple large language models, it enhances annotation accuracy through consensus-based predictions, reducing errors and providing uncertainty metrics. ### Key Features: - **Consensus-based Annotation**: Combines predictions from multiple large language models to improve accuracy. - **Integration with scRNA-seq Tools**: Works seamlessly with platforms like Scanpy and Seurat. - **No Reference Dataset Required**: Annotation does not require pre-existing reference datasets. - **Open-Source and Python-Friendly**: Available as an R package and Python package, making it accessible and easy to use. ### Tech Stack: Python ### Why It Matters: The integration of multiple models in mLLMCelltype offers a significant advantage over single-model approaches by reducing errors and providing a more reliable annotation. This framework is particularly useful in the field of single-cell transcriptomics, where accurate cell type identification is crucial for understanding cellular heterogeneity and disease mechanisms.

Overview: A fast, accurate, and easy-to-run Python wrapper for dense optical flow, built on Ce Liu's C++ implementation of the Coarse2Fine method. Key features: - Fast and accurate optical flow method - Based on the Coarse2Fine warping technique from Thomas Brox - Minimal dependencies and eliminates the need for OpenCV - Real-time performance optimization available with image resizing Tech stack: C++ Why it matters: This Python wrapper is designed for real-time performance and ease of use, making it ideal for applications requiring fast and accurate optical flow analysis without the need for complex libraries like OpenCV.

Overview: AnyV2V is a framework designed for video-to-video editing tasks, specifically focusing on performing video editing with a single image. It leverages the AnyV2V model, which is a tuning-free approach that does not require any training or fine-tuning. This makes it highly efficient and versatile, capable of seamlessly integrating with existing image editing methods. Key features: - Perform Video Editing with a Single Image - Tuning-Free Training - Seamless Integration with Image Editing Methods Tech stack: The project utilizes Jupyter Notebook as the primary language for development and documentation. Why it matters: AnyV2V simplifies the video editing process by abstracting it into an image editing problem, making it easier to leverage existing image editing techniques. This framework is particularly useful for tasks such as image-based video editing, where it can be used to perform diverse types of video edits without the need for extensive training or fine-tuning. Repository: TIGER-AI-Lab/AnyV2V GitHub stars: 660 GitHub description: Code and data for "AnyV2V: A Tuning-Free Framework For Any Video-to-Video Editing Tasks" [TMLR 2024] [**🌐 Homepage**](https://tiger-ai-lab.github.io/AnyV2V/) | [**📖 arXiv**](https://arxiv.org/abs/2403.14468) | [**🤗 HuggingFace Demo**](https://huggingface.co/spaces/TIGER-L

### Overview: NOVA is an autoregressive video generation model that achieves high efficiency without using vector quantization. It reformulates the video generation problem as non-quantized autoregressive modeling for both temporal and spatial predictions. ### Key Features: - **Non-Quantized Autoregressive Generation**: NOVA generates videos without vector quantization, offering high efficiency. - **Unified Modeling**: The model can perform both text-to-image and image-to-video tasks, showcasing its versatility. - **State-of-the-Art Performance**: NOVA achieves state-of-the-art results in both text-to-image and image-to-video tasks. ### Tech Stack: Python ### Why It Matters: NOVA provides a novel approach to video generation that is both efficient and versatile, making it a significant advancement in the field of image and video synthesis.

### Overview: The Ego4d dataset repository is a comprehensive collection of large-scale multi-modal video data, including 3D video recordings, spanning over 3700 hours of first-person video data. It serves as a benchmark for video learning and perception tasks, providing essential tools and resources for researchers and developers in the field. ### Key Features: - **Ego-Exo4D Dataset**: A large-scale multi-modal multi-view video dataset with time-synchronized videos of participants recorded with egocentric (first-person) and exocentric (third-person) perspective cameras. - **Ego4D Dataset**: The world's largest egocentric video ML dataset, featuring over 3700 hours of annotated first-person video data. - **Data Access**: Provides access to the dataset through command-line tools and a simple API for reading videos. - **Feature Extraction**: Offers lightweight wrappers for common models, including lightweight wrappers for models like Omnivore and SlowFast, which are essential for researchers and developers to analyze and utilize the dataset. - **Examples and Tutorials**: Includes a set of notebooks for Ego4D and Ego-Exo4D datasets, serving as examples and tutorials to help users understand and utilize the dataset effectively. ### Tech Stack: The repository is built using Python, with the `ego4d` module abstracting common video reading libraries such as `TorchAudio` and `PyAV`. ### Why It Matters: The Ego4d dataset and repository are crucial for researchers and developers in the field of video learning and perception

Overview: Macropodus is a comprehensive Natural Language Processing (NLP) toolkit built on top of the Albert+BiLSTM+CRF deep learning network architecture. It includes features such as Chinese word segmentation, part-of-speech tagging, named entity recognition, new word discovery, keyword extraction, text summarization, text similarity, scientific calculator, Chinese number to Arabic number conversion, Chinese to Roman number conversion, and pinyin conversion. Key features: - Chinese word segmentation (cut_bidirectional) - Part-of-speech tagging (cut_forward) - Named entity recognition (cut_reverse) - New word discovery - Keyword extraction - Text summarization - Text similarity - Scientific calculator - Chinese number to Arabic number conversion - Chinese to Roman number conversion - Pinyin conversion Tech stack: Python Why it matters: Macropodus is a powerful tool for natural language processing tasks, providing essential functionalities such as text summarization, keyword extraction, and more, making it highly useful for applications requiring NLP capabilities. Repository: yongzhuo/Macropodus Primary language: Python GitHub stars: 660 GitHub description: Macropodus is a comprehensive Natural Language Processing (NLP) toolkit built on top of the Albert+BiLSTM+CRF deep learning network architecture. It includes features such as Chinese word segmentation, part-of-speech tagging, named entity recognition, new word discovery, keyword extraction, text summarization, text similarity, scientific calculator, Chinese number to Arabic number conversion, Chinese to Roman number conversion, and pinyin

### Overview We formalize the randomness in diffusion models as a latent space, which can be inferred from real images. This allows us to perform zero-shot image-to-image translation with text-to-image diffusion models and traditional unpaired image-to-image translation using diffusion models trained on related domains. ### Key Features - **CycleDiffusion**: A simple method capable of zero-shot image-to-image translation with text-to-image diffusion models and traditional unpaired image-to-image translation using diffusion models trained on related domains. - **Latent Space Formalization**: We formalize the randomness in diffusion models as a latent space, which can be inferred from real images. - **Real Image Input**: All source images in our experiments are real images, including those generated by DALL•E 2. ### Tech Stack - Python ### Why It Matters This repository contains the code for CycleDiffusion, a method that leverages the latent space of diffusion models to perform zero-shot image-to-image translation and traditional unpaired image-to-image translation.

### Overview: IndoNLU is a comprehensive benchmark and resource for Indonesian natural language processing (NLP) that includes 12 downstream tasks, pre-trained models, and a starter code. This project was presented at AACL-IJCNLP 2020 and is designed to evaluate the performance of Indonesian language models. ### Key features: - **12 Downstream Tasks**: Covering various NLP tasks such as Named Entity Recognition, Part-of-Speech Tagging, Sentence Classification, and more. - **IndoBERT Models**: Large pre-trained models trained on a 4 billion word corpus, including two versions: IndoBERT-base and IndoBERT-lite. - **Indo4B Dataset**: A large pretraining dataset of 23 GB uncompressed and 5.6 GB compressed, used for model training and evaluation. - **GitHub Repository**: The project is hosted on GitHub, with 659 stars, making it a popular resource in the NLP community. - **Documentation**: Comprehensive documentation including contributing guidelines, examples, and submission instructions. ### Tech stack: Jupyter Notebook ### Why it matters: IndoNLU serves as a crucial benchmark for evaluating Indonesian language models, providing a standardized environment for researchers and practitioners to test and compare their models. Its inclusion of various downstream tasks and pre-trained models makes it an essential tool for advancing NLP research in Indonesia. ### Repository: IndoNLU - **Primary Language**: Jupyter Notebook - **GitHub Stars**: 659 - **GitHub Description**: The first-ever vast natural language processing

Overview: Polywise is an open-source agentic content system designed for self-hosted use, allowing users to interact with models, save knowledge, and retrieve context through a command-line interface or desktop application. Key features: - Command-line interface (CLI) for quick access - Desktop application for a seamless user experience - Model providers and integration setup - Embedding and rerank models for saved-content retrieval - Web UI for easy access to API - Authentication login for secure access Tech stack: TypeScript Why it matters: Polywise offers a flexible and accessible solution for creating and managing agentic content, making it ideal for developers looking to integrate AI into their projects without the need for a dedicated server. Repository: MatrixAges/polywise GitHub stars: 659

### Overview: EMA-Pytorch is a simple library designed to help developers keep track of an Exponential Moving Average (EMA) version of their PyTorch model. This allows for better performance monitoring and optimization of their models over time. ### Key Features: - **Exponential Moving Average (EMA) Tracking**: Automatically updates the EMA of your PyTorch model's weights. - **Post-Hoc Synthesis of EMA**: A method to create a new EMA model based on previously saved checkpoints, useful for post-hoc analysis. - **Target Representation Routing**: Automatically applies EMA outputs to the forward pass of nested module trees, such as those used in self-supervised learning. ### Tech Stack: Python ### Why It Matters: EMA-Pytorch simplifies the process of tracking and optimizing your PyTorch models by automatically updating their EMA versions. This can lead to better performance monitoring and optimization, making it easier to fine-tune your models over time.

### Overview: Awesome-LLM-Eval is a curated list of tools, datasets, benchmarks, demos, leaderboards, papers, and models primarily aimed at evaluating large language models (LLMs) and exploring the boundaries and limits of generative AI. ### Key features: - A comprehensive collection of tools, datasets, benchmarks, demos, leaderboards, papers, and models. - Focus on evaluation of LLMs, including various categories such as general, domain-specific, RAG-Evaluation, agent capabilities, coding capabilities, multimodal/cross-modal, long-context, inference speed, and quantization/compres - Updated regularly to reflect the latest developments in the field. ### Tech stack: Python, JavaScript, Markdown, and various open-source tools/libraries. ### Why it matters: This project serves as a valuable resource for researchers, developers, and practitioners interested in understanding and advancing the capabilities and limitations of large language models. By providing a curated list of resources, it facilitates the exploration of new evaluation methods and benchmarks, enabling deeper insights into the performance and potential of these models.

### Overview: This repository is a curated list of large AI models, or foundation models, designed for reasoning tasks. It covers three categories: language, vision, and multimodal models, and discusses various reasoning applications including commonsense, mathematical, logical, causal, visual, audio, multimodal, and agent reasoning. ### Key Features: - Comprehensive list of large AI models for reasoning. - Organized into language, vision, and multimodal categories. - Detailed elaboration on reasoning tasks such as commonsense, mathematical, logical, causal, visual, audio, multimodal, and agent reasoning. - Summarizes reasoning techniques like pre-training, fine-tuning, alignment training, mixture of experts, in-context learning, and autonomous agent. ### Tech Stack: - Python, Jupyter Notebook, Markdown, LaTeX. ### Why it Matters: This repository provides a valuable resource for researchers and developers looking to explore and leverage foundation models for reasoning tasks. It serves as a comprehensive guide to the current state of the art in reasoning with large AI models, making it an essential tool for anyone interested in advancing the field of artificial intelligence.