Coding & Development

attention_with_linear_biases is a GitHub repository offering the implementation of the Attention with Linear Biases (ALiBi) method for transformer language models. This method, presented in the ICLR 2022 paper 'Train Short, Test Long,' allows models to be trained on shorter input sequences (e.g., 1024 tokens) and then perform inference on significantly longer sequences (e.g., 2048 tokens or more) without requiring fine-tuning. The repository provides code and models for conducting experiments, specifically on the WikiText-103 dataset. ALiBi simplifies the positional encoding process by adding a linear bias to each attention score instead of using traditional position embeddings, which can improve performance even in non-extrapolating scenarios. The implementation details, including removing position embeddings and setting up the relative bias matrix, are clearly outlined.

attorch offers a collection of PyTorch's neural network modules, re-implemented in Python using OpenAI's Triton. The project's core goal is to provide an easily hackable, self-contained, and readable set of deep learning operations, maintaining or improving efficiency compared to standard PyTorch implementations. It serves as an accessible starting point for developers looking to create custom deep learning operations without the speed limitations of pure PyTorch or the complexity of writing CUDA kernels. Unlike many Triton-powered frameworks focused on Transformers, attorch includes layers for diverse applications like computer vision. It supports both forward and backward passes, making it suitable for training and inference, and offers an interface with PyTorch fallback for seamless integration.

awesome-agents is a comprehensive, curated list of open-source tools and products designed for building AI agents. This resource is invaluable for developers and researchers looking to explore and implement AI agent technology. It categorizes tools into various sections, including Frameworks, Testing and Evaluation, Software Development, Research, Conversational/General Agents, Game/Simulation, Knowledge Management, Automation, Browser, and Multimodal. The list features prominent frameworks like LangChain, AutoGen, and CrewAI, alongside specialized tools for testing, code generation, and research. It serves as a central hub for discovering cutting-edge solutions and fostering collaboration within the AI agent development community.

awesome-assistants offers a curated, open-source collection of AI assistants designed to streamline daily tasks. This comprehensive list serves as a foundation for building packages across various programming languages, facilitating easy integration into diverse applications. Users can explore a wide range of assistants, from general-purpose helpers to specialized roles like marketing, coding, and financial advisors. The project also provides a Telegram bot for convenient testing of these AI assistants, leveraging the OpenAI API. It's an invaluable resource for developers and businesses looking to quickly implement and experiment with AI-powered functionalities.

awesome-graph-classification is a comprehensive collection of graph classification methods, encompassing embedding, deep learning, graph kernel, and factorization papers. This resource provides researchers and practitioners with a curated list of important papers, often accompanied by their reference implementations. It serves as a valuable starting point for exploring various techniques in graph-based machine learning, offering insights into areas like network embedding, graph convolutional networks, and graph attention networks. The repository also links to relevant graph classification benchmark datasets, making it a practical tool for academic research and development in the field.

awesome-uncertainty-deeplearning is an extensive open-source repository dedicated to predictive uncertainty estimation in deep learning models. It compiles a wide range of resources including surveys, academic papers, datasets, and code implementations. The collection covers various methodologies such as Bayesian methods, ensemble techniques, sampling/dropout-based approaches, post-hoc methods, data augmentation, and evidential deep learning. It also addresses applications in classification, regression, object detection, natural language processing, and more. This repository is an invaluable resource for researchers and practitioners looking to explore, understand, and implement uncertainty quantification in their deep learning projects.

Awesome-diffusion-model-for-image-processing is a comprehensive, open-source GitHub repository that serves as a summary of diffusion model-based image processing techniques. It covers a wide array of applications such as image restoration, enhancement, coding, and quality assessment. The repository is continuously updated with new related works and includes detailed sections on image super-resolution, video restoration, inpainting, denoising, dehazing, deblurring, and medical image restoration. It also features benchmarks, datasets, and models for image/video compression and quality assessment, making it an invaluable resource for researchers and practitioners in the field.

Awesome-CV-MasterHub is an open-source repository providing a curated list of recent Computer Vision (CV) papers. It serves as a valuable resource for researchers and practitioners looking to stay abreast of the latest developments in the field. The platform organizes papers by various CV sub-domains such as Image Classification, Object Detection, Semantic Segmentation, Image Generation, and Vision-LLMs. Users can easily browse through the list and find links to papers, with code links provided where available. The repository is actively maintained, with updates to ensure the most recent and relevant articles are included, typically retaining up to 200 papers per area. It encourages community contributions through issues and pull requests for any overlooked papers.

Automatic_Speech_Recognition is an open-source, end-to-end automatic speech recognition system built with TensorFlow. It provides comprehensive support for both Mandarin and English, enabling users to develop and fine-tune their own speech recognition models. The tool includes various acoustic modeling techniques such as RNN, BRNN, LSTM, BLSTM, GRU, BGRU, Dynamic RNN, and Deep Residual Networks. It also features Seq2Seq with attention decoder, CTC decoding, and robust data preprocessing for TIMIT and LibriSpeech corpora. Users can train models with CPU/GPU, manage logging, and leverage features like dropout for dynamic RNNs and shell script execution.

The aws-machine-learning-university-accelerated-cv repository offers comprehensive educational materials for the Machine Learning University (MLU) Computer Vision class. This open-source resource is designed to make machine learning accessible to everyone, providing a structured path to learn about widely used ML techniques and apply them to real-world problems in computer vision. The class includes three lectures covering topics such as Intro to Computer Vision, Neural Networks, Convolutional Neural Networks, Image Datasets, and advanced CNN architectures like VGGNet and ResNet. It also features a final project where students practice working with a real-world computer vision dataset. The repository contains slides, Jupyter notebooks for hands-on practice, and datasets, making it a valuable tool for self-paced learning and experimentation.

awesome-claude-code is a meticulously curated collection of resources designed to enhance the Claude Code workflow. This open-source repository features a wide array of skills, hooks, slash-commands, agent orchestrators, applications, and plugins specifically tailored for Anthropic's Claude Code. It includes tools for various development needs, from agent skills for specialized tasks like workflow automation and security auditing to comprehensive workflows and knowledge guides for project management and documentation. The list also covers IDE integrations, usage monitors, status lines, and version control tools, making it an invaluable resource for developers looking to optimize their use of Claude Code.

CatVTON is an innovative virtual try-on diffusion model designed for efficiency and accessibility. It boasts a lightweight network with 899.06M total parameters and parameter-efficient training, utilizing only 49.57M trainable parameters. This optimization allows for simplified inference, requiring less than 8GB VRAM for high-resolution outputs of 1024x768. CatVTON supports deployment via Gradio App and ComfyUI, with automatic checkpoint downloads from HuggingFace. It also provides evaluation code for calculating metrics on datasets like VITON-HD and DressCode, making it a comprehensive solution for virtual try-on research and application development. The project is open-source and was accepted to ICLR 2025.

Brilliant Labs is dedicated to fostering an open-source ecosystem, providing resources and tools for developers and creatives to innovate and shape the future. Their flagship product, Halo, is an open-source glasses platform designed for curious and creative individuals. Halo features a color microOLED display, bone conduction speakers, and an ultra low-power Alif B1 processor with a NPU for on-device AI. It includes an optical sensor for AI inference, microphones with audio activity detection, and a 6-axis IMU. Running on ZephyrOS with a Lua interface, Halo offers cross-platform mobile app connectivity and a cloud-based AI agent named Noa, which handles real-time, multimodal conversations and remembers past interactions to personalize experiences.

bert-extractive-summarizer is an open-source Python library designed for extractive text summarization, building upon the HuggingFace Pytorch transformers library. The tool operates by first embedding sentences from the input text and then employing a clustering algorithm to identify and extract sentences closest to the cluster centroids, forming a concise summary. It also incorporates coreference resolution techniques, utilizing the neuralcoref library, to enhance the coherence and context of the generated summaries. Users can customize various parameters, including the number of sentences or ratio for the summary, and integrate custom models or Sentence-BERT for diverse summarization needs. The library supports GPU acceleration via CUDA by default if available, and offers a Flask service with Docker support for easy deployment.

CCSR is an open-source tool designed to enhance image quality through content-consistent super-resolution, leveraging diffusion models. It provides official code for both CCSRv1 and the upgraded CCSRv2, which is built on Diffusers. CCSRv2 introduces significant improvements, including flexible diffusion step selection without retraining, allowing users to adjust steps to their specific needs. It boasts high efficiency, supporting inference with as few as 1 or 2 diffusion steps, drastically reducing computation time. The tool also delivers enhanced clarity with crisper details and improved stability in synthesizing fine image details, ensuring higher-quality outputs. CCSR streamlines the restoration process with a one-step diffusion workflow in its second stage.

Change-Detection-Review is an open-source resource offering a detailed review of artificial intelligence-based change detection methods, particularly within the domain of remote sensing. This GitHub repository compiles available codes and open datasets essential for deep learning applications in this field. It is based on the paper "Change detection based on artificial intelligence: state-of-the-art and challenges," providing insights into the implementation processes, data types (optical RS, SAR, street view, heterogeneous data), and general frameworks of AI-based change detection. The review also covers commonly used networks, application domains, and discusses major challenges and future prospects, making it a valuable resource for researchers.

Baichuan-7B is a large-scale 7B parameter pre-training language model developed by BaiChuan-Inc. Based on the Transformer structure, it was trained on approximately 1.2 trillion tokens and supports both Chinese and English languages. The model features a context window length of 4096 and has demonstrated strong performance on standard Chinese and English benchmarks like C-Eval and MMLU. It includes optimizations for training stability and throughput, such as efficient operators, operator splitting, mixed precision, and communication optimizations, achieving high GPU peak compute utilization. The model also features an optimized tokenizer for Chinese language compression and improved mathematical capabilities.

ContextGem is a free, open-source LLM framework designed to radically simplify the extraction of structured data and insights from various documents. It eliminates extensive boilerplate code often required by other frameworks, significantly reducing development time and complexity. Key features include automated dynamic prompts, data modeling and validators, precise granular reference mapping, and multilingual support. ContextGem allows users to extract structured data, identify key aspects, and build complex extraction workflows through an intuitive API. It supports both cloud-based and local LLMs via LiteLLM integration and offers optimizations for accuracy, speed, and cost, making it ideal for in-depth single-document analysis.

Leaderboard LLM FR is an AI tool hosted on Hugging Face, designed to track, rank, and evaluate open-source Large Language Models (LLMs) and chatbots, with a specific focus on French language capabilities. It enables users to compare the performance of various models across multiple benchmarks, including IFEval, BBH, MATH, GPQA, MUSR, and MMLU-PRO. The platform offers real-time filtering and voting functionalities, providing comprehensive insights into model performance. This tool is ideal for researchers, developers, and data scientists interested in benchmarking and understanding the landscape of open-source LLMs, particularly those with a focus on the French language.

Computer-vision-course is a comprehensive, community-led course designed to teach Computer Vision with Neural Networks. Developed by over 60 contributors from the Hugging Face Computer Vision community, this course offers a unique and diverse learning experience. It covers a wide range of topics including fundamentals, Convolutional Neural Networks (CNNs), Vision Transformers, Multimodal Models, Generative Models, Basic CV Tasks, Video and Video Processing, 3D Vision, Scene Rendering and Reconstruction, Model Optimization, Synthetic Data Creation, Zero Shot Computer Vision, and Ethics and Biases. The course emphasizes a community-powered approach, allowing authors freedom in their style while maintaining a structured curriculum. It's an excellent resource for anyone looking to deepen their understanding of computer vision.

cnn-lstm-bilstm-deepcnn-clstm-in-pytorch is an open-source project offering implementations of several neural network architectures within the PyTorch framework. Designed for classification tasks, it includes models such as Convolutional Neural Networks (CNN), Bidirectional Long Short-Term Memory (BiLSTM), Bi-GRU, and DeepCNN. The repository provides a structured environment for experimenting with these models, particularly for sequence modeling and text classification applications. It details requirements like PyTorch 1.0.1 and Python 3.6, and offers configuration options for usage. The project also includes pre-trained models and results for SST-1 and SST-2 datasets, making it a valuable resource for developers and researchers working on deep learning projects in PyTorch.

deep-anpr is an open-source project designed for automatic number plate recognition (ANPR) using neural networks. This tool is presented as an experimental project, ideal for developers and researchers who wish to explore and tinker with ANPR technology. It requires dependencies such as TensorFlow, OpenCV, and NumPy. Users can extract background images, generate test set images, train the model (GPU recommended), and detect number plates in images. The project is noted as incomplete and not yet suitable for practical, production-level ANPR systems, but offers a solid foundation for those looking to understand and contribute to the development of such systems.

Ko AgentBench is a platform designed to evaluate and rank large language models (LLMs) based on their performance in agentic tasks. Hosted on Hugging Face, this tool offers a clear leaderboard that allows users to compare different LLMs. A key feature is its multilingual interface, enabling users to switch between Korean and English views with a simple click, catering to a broader audience. This makes it an invaluable resource for researchers, developers, and anyone interested in understanding the capabilities and limitations of various LLMs in practical, agent-based applications. The platform aims to provide transparent and accessible benchmarking data for the AI community.

Deep-learning-in-cloud is a comprehensive open-source GitHub repository that serves as a curated list of deep learning cloud providers. It aims to assist users in identifying suitable cloud GPUs for training their machine learning models more efficiently and cost-effectively. The resource also includes a section dedicated to MLOps platforms, offering insights into tools that support the complete machine learning lifecycle, from development to deployment and management. Additionally, it provides information on deploying models as web applications and highlights various perks and offers, including free credits and programs for students, researchers, and startups.