ring-attention-pytorch

This repository contains an implementation of Ring Attention, a technique for processing large sequences in transformers. Ring Attention splits the data across the sequence dimension and applies ring reduce to the processing of the tiles of the attention matrix, similar to flash attention. It also includes support for Striped Attention, a follow-up paper that permutes the sequence for better workload balancing for autoregressive transformers, and grouped query attention, which saves on communication costs during the ring reduce. The repository includes a CUDA version of the flash attention kernel, which is used for the forward and backward passes of the ring attention. It also includes logic for splitting the sequence evenly among ranks, either within the attention function or in the external ring transformer wrapper, and basic test cases with two processes to check for equivalent output and gradients.

This repository contains the official implementation of RPG, a powerful training-free paradigm for text-to-image generation and editing. RPG utilizes proprietary or open-source MLLMs as prompt recaptioner and region planner with complementary regional diffusion. It achieves state-of-the-art results and can generate high-resolution images. The codebase supports diffusers and various diffusion backbones, including SDXL and SD v1.4/1.5. Users can reproduce results with GPT-4, Gemini-Pro, or local MLLMs like miniGPT-4. The repository provides tools for quick start, regional diffusion with GPT-4, and regional diffusion with local LLMs.

MiniCheck is an efficient fact-checking tool designed to verify claims against grounding documents using large language models. It provides a sentence-level fact-checking model that can be used to evaluate the consistency of claims with the provided documents. MiniCheck offers different models, including Bespoke-MiniCheck-7B, which is the state-of-the-art and commercially usable. The tool enables users to fact-check multi-sentence claims by breaking them down into individual sentences for optimal performance. It also supports automatic prefix caching for faster inference when repeatedly fact-checking the same document with different claims.

Gemini is an open-source model designed to handle multiple modalities such as text, audio, images, and videos. It utilizes a transformer architecture with special decoders for text and image generation. The model processes input sequences by transforming them into tokens and then decoding them to generate image outputs. Gemini differs from other models by directly feeding image embeddings into the transformer instead of using a visual transformer encoder. The model also includes a component called Codi for conditional generation. Gemini aims to effectively integrate image, audio, and video embeddings to enhance its performance.

Numpy.NET is the most complete .NET binding for NumPy, empowering .NET developers with extensive functionality for scientific computing, machine learning, and AI. It provides multi-dimensional arrays, matrices, linear algebra, FFT, and more via a strong typed API. Numpy.NET does not require a local Python installation, as it uses Python.Included to package embedded Python 3.7. Multi-threading must be handled carefully to avoid deadlocks or access violation exceptions. Performance considerations include overhead when calling NumPy from C# and the efficiency of data transfer between C# and Python. Numpy.NET aims to match the completeness of the original NumPy library and is generated using CodeMinion by parsing the NumPy documentation. The project is MIT licensed and supported by JetBrains.

Woodpecker is a tool designed to correct hallucinations in Multimodal Large Language Models (MLLMs) by introducing a training-free method that picks out and corrects inconsistencies between generated text and image content. It consists of five stages: key concept extraction, question formulation, visual knowledge validation, visual claim generation, and hallucination correction. Woodpecker can be easily integrated with different MLLMs and provides interpretable results by accessing intermediate outputs of the stages. The tool has shown significant improvements in accuracy over baseline models like MiniGPT-4 and mPLUG-Owl.

SheetCopilot is an assistant agent that manipulates spreadsheets by following user commands. It leverages Large Language Models (LLMs) to interact with spreadsheets like a human expert, enabling non-expert users to complete tasks on complex software such as Google Sheets and Excel via a language interface. The tool observes spreadsheet states, polishes generated solutions based on external action documents and error feedback, and aims to improve success rate and efficiency. SheetCopilot offers a dataset with diverse task categories and operations, supporting operations like entry & manipulation, management, formatting, charts, and pivot tables. Users can interact with SheetCopilot in Excel or Google Sheets, executing tasks like calculating revenue, creating pivot tables, and plotting charts. The tool's evaluation includes performance comparisons with leading LLMs and VBA-based methods on specific datasets, showcasing its capabilities in controlling various aspects of a spreadsheet.

RAG-FiT is a library designed to improve Language Models' ability to use external information by fine-tuning models on specially created RAG-augmented datasets. The library assists in creating training data, training models using parameter-efficient finetuning (PEFT), and evaluating performance using RAG-specific metrics. It is modular, customizable via configuration files, and facilitates fast prototyping and experimentation with various RAG settings and configurations.

RAG Foundry is a library designed to enhance Large Language Models (LLMs) by fine-tuning models on RAG-augmented datasets. It helps create training data, train models using parameter-efficient finetuning (PEFT), and measure performance using RAG-specific metrics. The library is modular, customizable using configuration files, and facilitates prototyping with various RAG settings and configurations for tasks like data processing, retrieval, training, inference, and evaluation.

FuseMedML is a Python framework designed to accelerate machine learning-based discovery in the medical field by promoting code reuse. It provides a flexible design concept where data is stored in a nested dictionary, allowing easy handling of multi-modality information. The framework includes components for creating custom models, loss functions, metrics, and data processing operators. Additionally, FuseMedML offers 'batteries included' key components such as fuse.data for data processing, fuse.eval for model evaluation, and fuse.dl for reusable deep learning components. It supports PyTorch and PyTorch Lightning libraries and encourages the creation of domain extensions for specific medical domains.

LLM Reasoners is a library that enables LLMs to conduct complex reasoning, with advanced reasoning algorithms. It approaches multi-step reasoning as planning and searches for the optimal reasoning chain, which achieves the best balance of exploration vs exploitation with the idea of "World Model" and "Reward". Given any reasoning problem, simply define the reward function and an optional world model (explained below), and let LLM reasoners take care of the rest, including Reasoning Algorithms, Visualization, LLM calling, and more!

This repository contains the code for HOLD, a method that jointly reconstructs hands and objects from monocular videos without assuming a pre-scanned object template. It can reconstruct 3D geometries of novel objects and hands, enabling template-free bimanual hand-object reconstruction, textureless object interaction with hands, and multiple objects interaction with hands. The repository provides instructions to download in-the-wild videos from HOLD, preprocess and train on custom videos, a volumetric rendering framework, a generalized codebase for single and two hand interaction with objects, a viewer to interact with predictions, and code to evaluate and compare with HOLD in HO3D. The repository also includes documentation for setup, training, evaluation, visualization, preprocessing custom sequences, and using HOLD on ARCTIC.

MInference is a tool designed to accelerate pre-filling for long-context Language Models (LLMs) by leveraging dynamic sparse attention. It achieves up to a 10x speedup for pre-filling on an A100 while maintaining accuracy. The tool supports various decoding LLMs, including LLaMA-style models and Phi models, and provides custom kernels for attention computation. MInference is useful for researchers and developers working with large-scale language models who aim to improve efficiency without compromising accuracy.

This repository contains code for training reward models for Deep Reinforcement Learning-based Reward-modulated Hierarchical Fine-tuning (DRL-based RLHF), Iterative Selection Fine-tuning (Rejection sampling fine-tuning), and iterative Decision Policy Optimization (DPO). The reward models are trained using a Bradley-Terry model based on the Gemma and Mistral language models. The resulting reward models achieve state-of-the-art performance on the RewardBench leaderboard for reward models with base models of up to 13B parameters.

MME-RealWorld is a benchmark designed to address real-world applications with practical relevance, featuring 13,366 high-resolution images and 29,429 annotations across 43 tasks. It aims to provide substantial recognition challenges and overcome common barriers in existing Multimodal Large Language Model benchmarks, such as small data scale, restricted data quality, and insufficient task difficulty. The dataset offers advantages in data scale, data quality, task difficulty, and real-world utility compared to existing benchmarks. It also includes a Chinese version with additional images and QA pairs focused on Chinese scenarios.

This repository contains an implementation of Ring Attention, a technique for processing large sequences in transformers. Ring Attention splits the data across the sequence dimension and applies ring reduce to the processing of the tiles of the attention matrix, similar to flash attention. It also includes support for Striped Attention, a follow-up paper that permutes the sequence for better workload balancing for autoregressive transformers, and grouped query attention, which saves on communication costs during the ring reduce. The repository includes a CUDA version of the flash attention kernel, which is used for the forward and backward passes of the ring attention. It also includes logic for splitting the sequence evenly among ranks, either within the attention function or in the external ring transformer wrapper, and basic test cases with two processes to check for equivalent output and gradients.

This repository provides a set of ROS 2 packages to integrate llama.cpp into ROS 2. By using the llama_ros packages, you can easily incorporate the powerful optimization capabilities of llama.cpp into your ROS 2 projects by running GGUF-based LLMs and VLMs.

LMDeploy is a toolkit for compressing, deploying, and serving LLM, developed by the MMRazor and MMDeploy teams. It has the following core features: * **Efficient Inference** : LMDeploy delivers up to 1.8x higher request throughput than vLLM, by introducing key features like persistent batch(a.k.a. continuous batching), blocked KV cache, dynamic split&fuse, tensor parallelism, high-performance CUDA kernels and so on. * **Effective Quantization** : LMDeploy supports weight-only and k/v quantization, and the 4-bit inference performance is 2.4x higher than FP16. The quantization quality has been confirmed via OpenCompass evaluation. * **Effortless Distribution Server** : Leveraging the request distribution service, LMDeploy facilitates an easy and efficient deployment of multi-model services across multiple machines and cards. * **Interactive Inference Mode** : By caching the k/v of attention during multi-round dialogue processes, the engine remembers dialogue history, thus avoiding repetitive processing of historical sessions.

This repository contains an implementation of Ring Attention, a technique for processing large sequences in transformers. Ring Attention splits the data across the sequence dimension and applies ring reduce to the processing of the tiles of the attention matrix, similar to flash attention. It also includes support for Striped Attention, a follow-up paper that permutes the sequence for better workload balancing for autoregressive transformers, and grouped query attention, which saves on communication costs during the ring reduce. The repository includes a CUDA version of the flash attention kernel, which is used for the forward and backward passes of the ring attention. It also includes logic for splitting the sequence evenly among ranks, either within the attention function or in the external ring transformer wrapper, and basic test cases with two processes to check for equivalent output and gradients.

The collection of pre-trained, state-of-the-art AI models. ailia SDK is a self-contained, cross-platform, high-speed inference SDK for AI. The ailia SDK provides a consistent C++ API across Windows, Mac, Linux, iOS, Android, Jetson, and Raspberry Pi platforms. It also supports Unity (C#), Python, Rust, Flutter(Dart) and JNI for efficient AI implementation. The ailia SDK makes extensive use of the GPU through Vulkan and Metal to enable accelerated computing. # Supported models 323 models as of April 8th, 2024

Curated Transformers is a transformer library for PyTorch that provides state-of-the-art models composed of reusable components. It supports various transformer architectures, including encoders like ALBERT, BERT, and RoBERTa, and decoders like Falcon, Llama, and MPT. The library emphasizes consistent type annotations, minimal dependencies, and ease of use for education and research. It has been production-tested by Explosion and will be the default transformer implementation in spaCy 3.7.

**bert4torch** is a high-level framework for training and deploying transformer models in PyTorch. It provides a simple and efficient API for building, training, and evaluating transformer models, and supports a wide range of pre-trained models, including BERT, RoBERTa, ALBERT, XLNet, and GPT-2. bert4torch also includes a number of useful features, such as data loading, tokenization, and model evaluation. It is a powerful and versatile tool for natural language processing tasks.

FlagEmbedding focuses on retrieval-augmented LLMs, consisting of the following projects currently: * **Long-Context LLM** : Activation Beacon * **Fine-tuning of LM** : LM-Cocktail * **Embedding Model** : Visualized-BGE, BGE-M3, LLM Embedder, BGE Embedding * **Reranker Model** : llm rerankers, BGE Reranker * **Benchmark** : C-MTEB

Eden AI aims to simplify the use and deployment of AI technologies by providing a unique API that connects to all the best AI engines. With the rise of **AI as a Service** , a lot of companies provide off-the-shelf trained models that you can access directly through an API. These companies are either the tech giants (Google, Microsoft , Amazon) or other smaller, more specialized companies, and there are hundreds of them. Some of the most known are : DeepL (translation), OpenAI (text and image analysis), AssemblyAI (speech analysis). There are **hundreds of companies** doing that. We're regrouping the best ones **in one place** !