Coding & Development

Deep-Learning-Paper-Review-and-Practice is an open-source GitHub repository dedicated to providing comprehensive reviews and practical code implementations for deep learning papers. The repository curates a selection of recent and highly influential deep learning research, categorized into areas such as Image Recognition, Natural Language Processing, Generative Models & Super Resolution, Modeling & Optimization, and Adversarial Examples & Backdoor Attacks. Each paper entry includes links to the original paper, a video review, a summary PDF, and corresponding code practices, making it an invaluable resource for understanding and applying cutting-edge deep learning techniques. Users can engage with the content by exploring detailed explanations and hands-on coding examples, fostering a deeper understanding of complex AI concepts.

Deep Architecture Genealogy is an open-source project dedicated to mapping the vast and rapidly evolving landscape of deep learning architectures. It provides a comprehensive genealogy, illustrating the relationships and progression of various models such as CNNs (AlexNet, VggNet, ResNet), Generative Models (GANs, VAEs), Reinforcement Learning Algorithms (A3C, DARLA), and RNNs (LSTM, GRU, Transformer). The project is community-maintained, encouraging contributions via pull requests to its text-based genealogy file. This resource is invaluable for researchers, students, and practitioners seeking to understand the historical development and interconnections of deep learning models, offering both a visual mindmap and a detailed text version of the architectural lineage.

DEvol (DeepEvolution) is an open-source project designed as a proof of concept for genetic neural architecture search within the Keras framework. It allows for the evolution of neural network structures, including convolutional and dense layers, by varying parameters such as feature maps, activation functions, dropout rates, batch normalization, and max pooling. While currently tailored for classification problems, its architecture can be extended to other output types. The tool demonstrates how genetic algorithms can optimize neural network design, achieving competitive accuracy on datasets like MNIST. It emphasizes the potential for parallel training, early stopping, and parameter selection to manage the computational complexity inherent in evolving numerous models.

The Low-bit Quantized Open LLM Leaderboard, hosted on Hugging Face by Intel, provides a comprehensive platform for tracking, ranking, and evaluating open large language models (LLMs) and chatbots. Users can easily search and filter models based on various criteria such as type, size, and precision, making it a valuable resource for comparing performance. This tool is particularly useful for AI researchers, developers, and data scientists who need to stay updated on the latest advancements in open-source LLMs and assess their suitability for different applications. Its searchable interface simplifies the process of identifying and analyzing models, contributing to more informed decision-making in AI development.

D's Machine Learning (dml) is an open-source machine learning toolkit for Python, built upon the robust foundations of NumPy and SciPy. It emphasizes both the correctness of its algorithms and computational efficiency. The toolkit includes a comprehensive set of machine learning implementations, such as Neural Networks, Logistic Regression (softmax), Decision Trees (CART algorithm), and various clustering algorithms including k-means, k-medoids, spectral clustering, and hierarchical clustering. Additionally, it features Adaboost, k-Nearest Neighbor (with kd-tree BBF), Naive Bayesian (supporting continuous and discrete features), Support Vector Machines, simple Convolutional Neural Networks, and Collaborative Filtering algorithms. The project currently supports Python 2, with a note that Python 3 users are not yet supported.

The Llm Contamination Detector is a specialized tool hosted on Hugging Face Spaces, developed by Yeyito, aimed at identifying potential contamination within large language models (LLMs). This tool is crucial for maintaining the integrity and reliability of AI models, particularly in research and development environments. By detecting contamination, it helps ensure that LLMs are trained on clean, unbiased datasets, leading to more accurate and trustworthy AI outputs. While the live website currently indicates a runtime error, its intended purpose is to provide a valuable resource for AI researchers and machine learning engineers who need to validate the quality and purity of their language models. The tool's availability on Hugging Face Spaces suggests an open and community-driven approach to AI development.

DeepIE is an open-source deep learning framework specifically designed for information extraction tasks. It offers comprehensive implementations and benchmarks for key natural language processing (NLP) challenges such as named entity recognition (NER), relation extraction, and event extraction. The repository includes detailed performance metrics across various datasets and methods, including comparisons with BERT and other lexicon-enhanced models. DeepIE also provides resources like curated lists of relevant research papers and information extraction competitions, making it a valuable tool for researchers and developers working on advanced NLP applications.

Make Custom Voices With KokoroTTS is a web-based tool hosted on Hugging Face Spaces, designed for creating unique voice profiles. It enables users to select from several pre-made voices, fine-tune their individual strengths using intuitive sliders, and then blend them together to form a single, custom voice. Once a custom voice is created, users can input any text, and the application will read it aloud using their newly mixed voice. This tool is ideal for experimenting with voice synthesis and exploring different vocal textures and tones.

exBERT is an open-source visual analysis tool designed to help users explore and understand the learned attention weights and contextual representations within various Hugging Face Transformer models. It supports models like BERT, GPT2, DistilBERT, and ALBERT. Users can input sentences, visualize attention patterns as curved lines, and search embeddings across an annotated corpus. Key features include toggling visibility of attention to [CLS] and [SEP] tokens, interactively masking tokens to observe attention changes, and viewing model predictions. It also allows searching for contextual representations of tokens across layers and discovering linguistic features learned by specific heads, making it an invaluable resource for AI researchers and NLP developers.

Evaluate is a comprehensive open-source library designed to simplify the evaluation and comparison of machine learning models and datasets. It offers a wide array of pre-implemented metrics covering tasks from Natural Language Processing to Computer Vision, including dataset-specific metrics. Users can easily load and apply these metrics across different ML frameworks like NumPy, Pandas, PyTorch, TensorFlow, and JAX. The library also facilitates comparisons between models and provides tools for dataset evaluation. A key feature is the ability to add new evaluation modules to the Hugging Face Hub, fostering community collaboration and allowing users to share and discover custom metrics. Evaluate includes type checking for inputs, detailed metric cards explaining usage and limitations, and supports community-driven metric development.

DualPipe is an innovative open-source algorithm designed for efficient training of large language models, specifically DeepSeek V3/R1. It introduces a bidirectional pipeline parallelism approach that fully overlaps forward and backward computation-communication phases, significantly reducing pipeline bubbles. This optimization leads to more efficient use of computational resources during deep learning model training. The project also offers DualPipeV, a concise V-shape schedule derived from DualPipe, further enhancing efficiency. Developers can integrate DualPipe into their PyTorch 2.0+ projects, with examples provided for quick start. It's particularly useful for those working on large-scale model training where communication overhead is a critical factor.

EasyEdit is an open-source framework designed for knowledge editing in Large Language Models (LLMs). It offers an easy-to-use interface for developers to modify and control the information LLMs know and how they behave. The framework supports a wide array of knowledge editing techniques, ranging from updating internal parameters and introducing additional parameters (EasyEdit 1.0) to real-time steering during inference without retraining (EasyEdit 2.0). It integrates various steering methods and provides tools for evaluation, including a hierarchical benchmark called SteerEval. EasyEdit also incorporates support for unstructured long-form knowledge editing datasets and methods, making it a comprehensive solution for researchers and developers working on LLM controllability and refinement.

fastembed-rs is a Rust library designed for efficient generation of vector embeddings and reranking, crucial for building advanced AI applications. It offers synchronous usage and leverages @pykeio/ort for high-performance ONNX inference and @huggingface/tokenizers for rapid encodings. The library supports a wide array of pre-trained models for text embeddings (including BGE, MiniLM, MPNet, Nomic, and Qwen3 models), sparse text embeddings (like Splade_PP_en_v1 and BGE-m3), and image embeddings (such as CLIP, ResNet50, and Unicom models). Additionally, it provides reranking functionalities with models like BGE-reranker-base. Developers can easily integrate fastembed-rs into their Rust projects, with options for custom model initialization and support for DirectML on Windows for GPU acceleration.

The Model Memory Calculator, developed by TitanML, is a valuable tool hosted on Hugging Face Spaces designed to assist developers and data scientists in understanding the memory footprint of their AI models. By providing an estimate of the resources needed to run specific models, it facilitates better planning and optimization of AI deployments. This calculator is particularly useful for those working with large language models or complex neural networks, where memory management is crucial for efficient operation and cost-effectiveness. It helps users make informed decisions about hardware requirements and model architecture, ensuring their AI projects are both performant and resource-aware.

Dromedary is an open-source, self-aligned language model developed by IBM, designed to create helpful, ethical, and reliable Large Language Models (LLMs) with minimal human supervision. It features a principle-driven self-alignment process, including the updated Dromedary-2 (SFT) which utilizes diverse user prompts and an improved prompt structure for enhanced performance. Dromedary-2 also introduces the SALMON (Self-ALignMent with principle-fOllowiNg reward models) training pipeline for RLAIF. The project provides model weights as delta weights for LLaMA, synthetic data for self-alignment, and a full training pipeline for reproduction, making it a valuable resource for researchers and developers in the NLP domain.

All-in-One Demo is an AI demonstration tool hosted on Hugging Face Spaces, designed to showcase various AI functionalities. It is built using Gradio, an open-source Python library for creating easy-to-use UI components for machine learning models. This tool is intended for individuals, developers, and researchers who wish to explore and test different AI models and applications. While the live website indicates a runtime error, suggesting it may not be currently operational, its purpose is to provide a platform for interacting with AI models. It is licensed under AFL-3.0, making it accessible for free use and modification.

Generative_Deep_Learning_2nd_Edition is the official code repository for the second edition of the O'Reilly book "Generative Deep Learning: Teaching Machines to Paint, Write, Compose and Play." This open-source resource provides practical code examples and outlines corresponding to the book's chapters, covering topics such as Variational Autoencoders, Generative Adversarial Networks, Autoregressive Models, Normalizing Flows, Energy-Based Models, Diffusion Models, Transformers, and advanced GANs. It is designed to help users learn and implement generative deep learning techniques, with instructions for setting up a Docker environment, downloading datasets, and using Tensorboard for monitoring experiments. The repository also includes guidance for using cloud virtual machines.

HEBO is an open-source library developed by Huawei Noah's Ark Lab, focusing on Bayesian optimization, reinforcement learning, and generative model research. It offers official implementations for a wide range of algorithms, including Heteroscedastic Evolutionary Bayesian Optimisation (HEBO), a framework for Combinatorial and Mixed-variable Bayesian Optimization (MCBO), and End-to-End Meta-Bayesian Optimisation with Transformer Neural Processes (NAP). The library also covers high-dimensional Bayesian optimization with random decompositions (RDUCB) and applications in antibody design (AntBO) and logic synthesis (BOiLS). Additionally, HEBO supports research in reinforcement learning, such as enhancing agents with local guides and safe reinforcement learning, and generative models like EM-LLM for episodic memory in LLMs. It serves as a comprehensive resource for researchers and developers in these advanced AI fields.

Hedwig is an open-source repository offering PyTorch deep learning models specifically designed for document classification tasks. Developed by the Data Systems Group at the University of Waterloo, it includes implementations of several prominent models such as DocBERT, Reg-LSTM, XML-CNN, HAN, Char-CNN, and Kim CNN. Each model directory contains a detailed README.md for further information. The project is designed for Python 3.6 and PyTorch 0.4, with clear instructions for environment setup using Anaconda and installation of dependencies. It also provides options for downloading necessary datasets like Reuters, AAPD, and IMDB, along with word2vec embeddings, making it a comprehensive resource for document classification research and application.

GPTeacher is a comprehensive collection of modular datasets, meticulously generated by GPT-4, designed to facilitate various AI training and development tasks. The collection includes several distinct datasets: General-Instruct, Roleplay-Instruct, Code-Instruct, and Toolformer. The General-Instruct dataset, comprising approximately 20,000 examples, focuses on diverse tasks such as Chain of Thought Reasoning, Logic Puzzles, and Wordplay. The Roleplay-Instruct dataset, now in its V2 (Supplemental) version, is 2.5 times larger than the original and features simulated conversations for character role-playing. The Code-Instruct dataset offers around 5,350 code task instructions across various programming languages. Additionally, the Toolformer dataset is designed for training models to use predefined tools like search, Python, and Wikipedia. All datasets are formatted to be compliant with Alpaca's dataset structure, including instruction, input, and output fields, making them easy to integrate into existing fine-tuning processes.

GPTQ provides an efficient, open-source implementation of the GPTQ algorithm for accurate post-training quantization of generative pretrained transformers. This tool enables developers to compress large language models from the OPT and BLOOM families down to 2, 3, or 4 bits, significantly reducing their memory footprint and computational requirements while maintaining accuracy. Key features include support for weight grouping, evaluation of perplexity on various language generation tasks, and performance evaluation on ZeroShot tasks. The repository also offers a 3-bit quantized matrix full-precision vector product CUDA kernel and benchmarking code for individual matrix-vector products and language generation with quantized models. Recent updates include static groups options, adjusted preprocessing for C4 and PTB, optimized 3-bit kernels for faster generation, and a minimal LLaMa integration with new tricks like `--act-order` and `--true-sequential` for improved accuracy.

Hunyuan-A13B is an innovative and open-source large language model (LLM) developed by Tencent Hunyuan, featuring a fine-grained Mixture-of-Experts (MoE) architecture. With 80 billion total parameters and only 13 billion active parameters, it delivers high performance while maintaining optimal resource efficiency. Key features include hybrid reasoning support with both fast and slow thinking modes, ultra-long context understanding up to 256K tokens, and enhanced agent capabilities. The model is optimized for efficient inference using Grouped Query Attention (GQA) and supports multiple quantization formats like FP8 and INT4, making it suitable for resource-constrained environments. It is ideal for researchers and developers seeking powerful yet computationally efficient AI solutions.

Heretic is an open-source tool designed for the fully automatic removal of censorship, also known as "safety alignment," from transformer-based language models. It achieves this without requiring expensive post-training processes, utilizing an advanced implementation of directional ablation combined with a TPE-based parameter optimizer powered by Optuna. This approach allows Heretic to automatically find high-quality ablation parameters by co-minimizing refusal rates and KL divergence from the original model, ensuring the decensored model retains as much original intelligence as possible. The tool supports most dense and many multimodal models, including various MoE architectures. It also offers research features for interpretability studies, such as plotting residual vectors and printing residual geometry details.

HostedGPT is a free, open-source alternative to ChatGPT, built as a Ruby on Rails application, allowing it to be hosted anywhere or run locally. It supports multiple AI providers including Anthropic, Google, Llama, and Groq, enabling users to switch assistants mid-conversation. The platform offers a polished interface with strong mobile support and German localization. Users only pay for their API usage from providers like OpenAI, Anthropic, and Google, as the HostedGPT app itself is free. It also helps users avoid common usage caps and provides features for collecting, searching, and sharing conversations across different providers. Deployment options include Render, Fly.io, Heroku, or self-hosting, with detailed instructions for each.