Best AI tools for< Inference On Devices >

Qualcomm AI Hub is a platform that allows users to run AI models on Snapdragon® 8 Elite devices. It provides a collaborative ecosystem for model makers, cloud providers, runtime, and SDK partners to deploy on-device AI solutions quickly and efficiently. Users can bring their own models, optimize for deployment, and access a variety of AI services and resources. The platform caters to various industries such as mobile, automotive, and IoT, offering a range of models and services for edge computing.

Anycores is an AI tool designed to optimize the performance of deep neural networks and reduce the cost of running AI models in the cloud. It offers a platform that provides automated solutions for tuning and inference consultation, optimized networks zoo, and platform for reducing AI model cost. Anycores focuses on faster execution, reducing inference time over 10x times, and footprint reduction during model deployment. It is device agnostic, supporting Nvidia, AMD GPUs, Intel, ARM, AMD CPUs, servers, and edge devices. The tool aims to provide highly optimized, low footprint networks tailored to specific deployment scenarios.

Luxonis is an AI application that offers Visual AI solutions engineered for precision edge inference. The application provides stereo depth cameras with unique features and quality, enabling users to perform advanced vision tasks on-device, reducing latency and bandwidth demands. With open-source DepthAI API, users can create and deploy custom vision solutions that scale with their needs. Luxonis also offers real-world training data for self-improving vision intelligence and operates flawlessly through vibrations, temperature shifts, and extended use. The application integrates advanced sensing capabilities with up to 48MP cameras, wide field of view, IMUs, microphones, ToF, thermal, IR illumination, and active stereo for unparalleled perception.

ONNX Runtime is a production-grade AI engine designed to accelerate machine learning training and inferencing in various technology stacks. It supports multiple languages and platforms, optimizing performance for CPU, GPU, and NPU hardware. ONNX Runtime powers AI in Microsoft products and is widely used in cloud, edge, web, and mobile applications. It also enables large model training and on-device training, offering state-of-the-art models for tasks like image synthesis and text generation.

Lamini is an enterprise-level LLM platform that offers precise recall with Memory Tuning, enabling teams to achieve over 95% accuracy even with large amounts of specific data. It guarantees JSON output and delivers massive throughput for inference. Lamini is designed to be deployed anywhere, including air-gapped environments, and supports training and inference on Nvidia or AMD GPUs. The platform is known for its factual LLMs and reengineered decoder that ensures 100% schema accuracy in the JSON output.

Roboflow is an AI tool designed for computer vision tasks, offering a platform that allows users to annotate, train, deploy, and perform inference on models. It provides integrations, ecosystem support, and features like notebooks, autodistillation, and supervision. Roboflow caters to various industries such as aerospace, agriculture, healthcare, finance, and more, with a focus on simplifying the development and deployment of computer vision models.

TextSynth is an AI tool that provides access to large language models such as Mistral, Llama, Stable Diffusion, Whisper for text-to-image, text-to-speech, and speech-to-text capabilities via a REST API and playground. It employs custom inference code for faster inference on standard GPUs and CPUs. Founded in 2020, TextSynth was among the first to offer access to the GPT-2 language model. The service is free with rate limitations, and users can opt for custom support at a small cost per request. All servers are located in France.

Wallaroo.AI is an AI inference platform that offers production-grade AI inference microservices optimized on OpenVINO for cloud and Edge AI application deployments on CPUs and GPUs. It provides hassle-free AI inferencing for any model, any hardware, anywhere, with ultrafast turnkey inference microservices. The platform enables users to deploy, manage, observe, and scale AI models effortlessly, reducing deployment costs and time-to-value significantly.

Segwise is an AI tool designed to help game developers increase their game's Lifetime Value (LTV) by providing insights into player behavior and metrics. The tool uses AI agents to detect causal LTV drivers, root causes of LTV drops, and opportunities for growth. Segwise offers features such as running causal inference models on player data, hyper-segmenting player data, and providing instant answers to questions about LTV metrics. It also promises seamless integrations with gaming data sources and warehouses, ensuring data ownership and transparent pricing. The tool aims to simplify the process of improving LTV for game developers.

Salad is a distributed GPU cloud platform that offers fully managed and massively scalable services for AI applications. It provides the lowest priced AI transcription in the market, with features like image generation, voice AI, computer vision, data collection, and batch processing. Salad democratizes cloud computing by leveraging consumer GPUs to deliver cost-effective AI/ML inference at scale. The platform is trusted by hundreds of machine learning and data science teams for its affordability, scalability, and ease of deployment.

Groq is a fast AI inference tool that offers GroqCloud™ Platform and GroqRack™ Cluster for developers to build and deploy AI models with ultra-low-latency inference. It provides instant intelligence for openly-available models like Llama 3.1 and is known for its speed and compatibility with other AI providers. Groq powers leading openly-available AI models and has gained recognition in the AI chip industry. The tool has received significant funding and valuation, positioning itself as a strong challenger to established players like Nvidia.

Tensoic AI is an AI tool designed for custom Large Language Models (LLMs) fine-tuning and inference. It offers ultra-fast fine-tuning and inference capabilities for enterprise-grade LLMs, with a focus on use case-specific tasks. The tool is efficient, cost-effective, and easy to use, enabling users to outperform general-purpose LLMs using synthetic data. Tensoic AI generates small, powerful models that can run on consumer-grade hardware, making it ideal for a wide range of applications.

Mystic.ai is an AI tool designed to deploy and scale Machine Learning models with ease. It offers a fully managed Kubernetes platform that runs in your own cloud, allowing users to deploy ML models in their own Azure/AWS/GCP account or in a shared GPU cluster. Mystic.ai provides cost optimizations, fast inference, simpler developer experience, and performance optimizations to ensure high-performance AI model serving. With features like pay-as-you-go API, cloud integration with AWS/Azure/GCP, and a beautiful dashboard, Mystic.ai simplifies the deployment and management of ML models for data scientists and AI engineers.

Thirdai.com is an AI tool that offers a robot challenge screen for checking site connection security. The tool helps users assess the security of their website by requiring cookies to be enabled in the browser settings. It ensures that the connection is secure and provides recommendations for improving security measures.

Cortex Labs is a decentralized world computer that enables AI and AI-powered decentralized applications (dApps) to run on the blockchain. It offers a Layer2 solution called ZkMatrix, which utilizes zkRollup technology to enhance transaction speed and reduce fees. Cortex Virtual Machine (CVM) supports on-chain AI inference using GPU, ensuring deterministic results across computing environments. Cortex also enables machine learning in smart contracts and dApps, fostering an open-source ecosystem for AI researchers and developers to share models. The platform aims to solve the challenge of on-chain machine learning execution efficiently and deterministically, providing tools and resources for developers to integrate AI into blockchain applications.

Nesa is a global blockchain network that brings AI on-chain, allowing applications and protocols to seamlessly integrate with AI. It offers secure execution for critical inference, a private AI network, and a global AI model repository. Nesa supports various AI models for tasks like text classification, content summarization, image generation, language translation, and more. The platform is backed by a team with extensive experience in AI and deep learning, with numerous awards and recognitions in the field.

d-Matrix is an AI tool that offers ultra-low latency batched inference for generative AI technology. It introduces Corsair™, the world's most efficient AI inference platform for datacenters, providing high performance, efficiency, and scalability for large-scale inference tasks. The tool aims to transform the economics of AI inference by delivering fast, sustainable, and scalable AI solutions without compromising on speed or usability.

fal.ai is a generative media platform designed for developers to build the next generation of creativity. It offers lightning-fast inference with no compromise on quality, providing access to high-quality generative media models optimized by the fal Inference Engine™. The platform allows developers to fine-tune their own models, leverage real-time infrastructure for new user experiences, and scale to thousands of GPUs as needed. With a focus on developer experience, fal.ai aims to be the fastest AI tool for running diffusion models.

FluidStack is a leading GPU cloud platform designed for AI and LLM (Large Language Model) training. It offers unlimited scale for AI training and inference, allowing users to access thousands of fully-interconnected GPUs on demand. Trusted by top AI startups, FluidStack aggregates GPU capacity from data centers worldwide, providing access to over 50,000 GPUs for accelerating training and inference. With 1000+ data centers across 50+ countries, FluidStack ensures reliable and efficient GPU cloud services at competitive prices.

FuriosaAI is an AI application that offers Hardware RNGD for LLM and Multimodality, as well as WARBOY for Computer Vision. It provides a comprehensive developer experience through the Furiosa SDK, Model Zoo, and Dev Support. The application focuses on efficient AI inference, high-performance LLM and multimodal deployment capabilities, and sustainable mass adoption of AI. FuriosaAI features the Tensor Contraction Processor architecture, software for streamlined LLM deployment, and a robust ecosystem support. It aims to deliver powerful and efficient deep learning acceleration while ensuring future-proof programmability and efficiency.

MNN is a highly efficient and lightweight deep learning framework that supports inference and training of deep learning models. It has industry-leading performance for on-device inference and training. MNN has been integrated into various Alibaba Inc. apps and is used in scenarios like live broadcast, short video capture, search recommendation, and product searching by image. It is also utilized on embedded devices such as IoT. MNN-LLM and MNN-Diffusion are specific runtime solutions developed based on the MNN engine for deploying language models and diffusion models locally on different platforms. The framework is optimized for devices, supports various neural networks, and offers high performance with optimized assembly code and GPU support. MNN is versatile, easy to use, and supports hybrid computing on multiple devices.

TPI-LLM (Tensor Parallelism Inference for Large Language Models) is a system designed to bring LLM functions to low-resource edge devices, addressing privacy concerns by enabling LLM inference on edge devices with limited resources. It leverages multiple edge devices for inference through tensor parallelism and a sliding window memory scheduler to minimize memory usage. TPI-LLM demonstrates significant improvements in TTFT and token latency compared to other models, and plans to support infinitely large models with low token latency in the future.

JetStream is a throughput and memory optimized engine for Large Language Model (LLM) inference on XLA devices, specifically TPUs. It provides reference engine implementations for Jax and Pytorch models, along with documentation for online inference, serving Gemma using TPUs on GKE, benchmarking, observability, profiling, and standalone local setup. Users can easily set up a local server, run tests, and test core modules. JetStream aims to enhance the performance of LLM inference on XLA devices.

JetStream is a throughput and memory optimized engine for LLM inference on XLA devices, starting with TPUs (and GPUs in future -- PRs welcome). It is designed to provide high performance and scalability for large language models, enabling efficient inference on cloud-based TPUs. JetStream leverages XLA to optimize the execution of LLM models, resulting in faster and more efficient inference. Additionally, JetStream supports quantization techniques to further enhance performance and reduce memory consumption. By utilizing JetStream, developers can deploy and run LLM models on TPUs with ease, achieving optimal performance and cost-effectiveness.

T-MAC is a kernel library that directly supports mixed-precision matrix multiplication without the need for dequantization by utilizing lookup tables. It aims to boost low-bit LLM inference on CPUs by offering support for various low-bit models. T-MAC achieves significant speedup compared to SOTA CPU low-bit framework (llama.cpp) and can even perform well on lower-end devices like Raspberry Pi 5. The tool demonstrates superior performance over existing low-bit GEMM kernels on CPU, reduces power consumption, and provides energy savings. It achieves comparable performance to CUDA GPU on certain tasks while delivering considerable power and energy savings. T-MAC's method involves using lookup tables to support mpGEMM and employs key techniques like precomputing partial sums, shift and accumulate operations, and utilizing tbl/pshuf instructions for fast table lookup.

Run your own AI cluster at home with everyday devices. Exo is experimental software that unifies existing devices into a powerful GPU, supporting wide model compatibility, dynamic model partitioning, automatic device discovery, ChatGPT-compatible API, and device equality. It does not use a master-worker architecture, allowing devices to connect peer-to-peer. Exo supports different partitioning strategies like ring memory weighted partitioning. Installation is recommended from source. Documentation includes example usage on multiple MacOS devices and information on inference engines and networking modules. Known issues include the iOS implementation lagging behind Python.

OpenVINO™ is an open-source toolkit for optimizing and deploying AI inference. It provides a common API to deliver inference solutions on various platforms, including CPU, GPU, NPU, and heterogeneous devices. OpenVINO™ supports pre-trained models from Open Model Zoo and popular frameworks like TensorFlow, PyTorch, and ONNX. Key components of OpenVINO™ include the OpenVINO™ Runtime, plugins for different hardware devices, frontends for reading models from native framework formats, and the OpenVINO Model Converter (OVC) for adjusting models for optimal execution on target devices.

The LlamaEdge project makes it easy to run LLM inference apps and create OpenAI-compatible API services for the Llama2 series of LLMs locally. It provides a Rust+Wasm stack for fast, portable, and secure LLM inference on heterogeneous edge devices. The project includes source code for text generation, chatbot, and API server applications, supporting all LLMs based on the llama2 framework in the GGUF format. LlamaEdge is committed to continuously testing and validating new open-source models and offers a list of supported models with download links and startup commands. It is cross-platform, supporting various OSes, CPUs, and GPUs, and provides troubleshooting tips for common errors.

SmolChat-Android is a mobile application that enables users to interact with local small language models (SLMs) on-device. Users can add/remove SLMs, modify system prompts and inference parameters, create downstream tasks, and generate responses. The app uses llama.cpp for model execution, ObjectBox for database storage, and Markwon for markdown rendering. It provides a simple, extensible codebase for on-device machine learning projects.

Welcome to the ultimate hub for on-device Large Language Models (LLMs)! This repository is your go-to resource for all things related to LLMs designed for on-device deployment. Whether you're a seasoned researcher, an innovative developer, or an enthusiastic learner, this comprehensive collection of cutting-edge knowledge is your gateway to understanding, leveraging, and contributing to the exciting world of on-device LLMs.

Distributed Llama is a tool that allows you to run large language models (LLMs) on weak devices or make powerful devices even more powerful by distributing the workload and dividing the RAM usage. It uses TCP sockets to synchronize the state of the neural network, and you can easily configure your AI cluster by using a home router. Distributed Llama supports models such as Llama 2 (7B, 13B, 70B) chat and non-chat versions, Llama 3, and Grok-1 (314B).

AWQ (Activation-aware Weight Quantization) is a tool designed for efficient and accurate low-bit weight quantization (INT3/4) for Large Language Models (LLMs). It supports instruction-tuned models and multi-modal LMs, providing features such as AWQ search for accurate quantization, pre-computed AWQ model zoo for various LLMs, memory-efficient 4-bit linear in PyTorch, and efficient CUDA kernel implementation for fast inference. The tool enables users to run large models on resource-constrained edge platforms, delivering more efficient responses with LLM/VLM chatbots through 4-bit inference.

KanTV is an open-source project that focuses on studying and practicing state-of-the-art AI technology in real applications and scenarios, such as online TV playback, transcription, translation, and video/audio recording. It is derived from the original ijkplayer project and includes many enhancements and new features, including: * Watching online TV and local media using a customized FFmpeg 6.1. * Recording online TV to automatically generate videos. * Studying ASR (Automatic Speech Recognition) using whisper.cpp. * Studying LLM (Large Language Model) using llama.cpp. * Studying SD (Text to Image by Stable Diffusion) using stablediffusion.cpp. * Generating real-time English subtitles for English online TV using whisper.cpp. * Running/experiencing LLM on Xiaomi 14 using llama.cpp. * Setting up a customized playlist and using the software to watch the content for R&D activity. * Refactoring the UI to be closer to a real commercial Android application (currently only supports English). Some goals of this project are: * To provide a well-maintained "workbench" for ASR researchers interested in practicing state-of-the-art AI technology in real scenarios on mobile devices (currently focusing on Android). * To provide a well-maintained "workbench" for LLM researchers interested in practicing state-of-the-art AI technology in real scenarios on mobile devices (currently focusing on Android). * To create an Android "turn-key project" for AI experts/researchers (who may not be familiar with regular Android software development) to focus on device-side AI R&D activity, where part of the AI R&D activity (algorithm improvement, model training, model generation, algorithm validation, model validation, performance benchmark, etc.) can be done very easily using Android Studio IDE and a powerful Android phone.

Awesome-LLM-Quantization is a curated list of resources related to quantization techniques for Large Language Models (LLMs). Quantization is a crucial step in deploying LLMs on resource-constrained devices, such as mobile phones or edge devices, by reducing the model's size and computational requirements.

MiniCPM-V is a series of end-side multimodal LLMs designed for vision-language understanding. The models take image and text inputs to provide high-quality text outputs. The series includes models like MiniCPM-Llama3-V 2.5 with 8B parameters surpassing proprietary models, and MiniCPM-V 2.0, a lighter model with 2B parameters. The models support over 30 languages, efficient deployment on end-side devices, and have strong OCR capabilities. They achieve state-of-the-art performance on various benchmarks and prevent hallucinations in text generation. The models can process high-resolution images efficiently and support multilingual capabilities.

Green-Bit-LLM is a Python toolkit designed for fine-tuning, inferencing, and evaluating GreenBitAI's low-bit Language Models (LLMs). It utilizes the Bitorch Engine for efficient operations on low-bit LLMs, enabling high-performance inference on various GPUs and supporting full-parameter fine-tuning using quantized LLMs. The toolkit also provides evaluation tools to validate model performance on benchmark datasets. Green-Bit-LLM is compatible with AutoGPTQ series of 4-bit quantization and compression models.

LLaMA Factory is a unified framework for fine-tuning 100+ large language models (LLMs) with various methods, including pre-training, supervised fine-tuning, reward modeling, PPO, DPO and ORPO. It features integrated algorithms like GaLore, BAdam, DoRA, LongLoRA, LLaMA Pro, LoRA+, LoftQ and Agent tuning, as well as practical tricks like FlashAttention-2, Unsloth, RoPE scaling, NEFTune and rsLoRA. LLaMA Factory provides experiment monitors like LlamaBoard, TensorBoard, Wandb, MLflow, etc., and supports faster inference with OpenAI-style API, Gradio UI and CLI with vLLM worker. Compared to ChatGLM's P-Tuning, LLaMA Factory's LoRA tuning offers up to 3.7 times faster training speed with a better Rouge score on the advertising text generation task. By leveraging 4-bit quantization technique, LLaMA Factory's QLoRA further improves the efficiency regarding the GPU memory.

Efficient Foundation Model Survey is a comprehensive analysis of resource-efficient large language models (LLMs) and multimodal foundation models. The survey covers algorithmic and systemic innovations to support the growth of large models in a scalable and environmentally sustainable way. It explores cutting-edge model architectures, training/serving algorithms, and practical system designs. The goal is to provide insights on tackling resource challenges posed by large foundation models and inspire future breakthroughs in the field.

Academic_LLM_Sec_Papers is a curated collection of academic papers related to LLM Security Application. The repository includes papers sorted by conference name and published year, covering topics such as large language models for blockchain security, software engineering, machine learning, and more. Developers and researchers are welcome to contribute additional published papers to the list. The repository also provides information on listed conferences and journals related to security, networking, software engineering, and cryptography. The papers cover a wide range of topics including privacy risks, ethical concerns, vulnerabilities, threat modeling, code analysis, fuzzing, and more.

SillyTavern is a user interface you can install on your computer (and Android phones) that allows you to interact with text generation AIs and chat/roleplay with characters you or the community create. SillyTavern is a fork of TavernAI 1.2.8 which is under more active development and has added many major features. At this point, they can be thought of as completely independent programs.

コードを分析し、言語ではなくスタイルに焦点を当て、プログラムを書いた人の性格を推察するツールです。( It is a tool that analyzes code, focuses on style rather than language, and infers the personality of the person who wrote the program. )

I provide detailed summaries, critiques, and inferences on articles, papers, transcripts, websites, and more. Just give me text, a URL, or file to digest.

Strategic advisor on power dynamics, inspired by 'The 48 Laws of Power'

Java & Python programming assistant for AI model inference.

Demystifying Experimentation and Causal Inference with 1-Sided Tests Focus

CUDA and Triton expert with concise, accurate answers.

Synthesizes insights from multiple branding books

Analyzes data to influence strategic decision-making processes.

Leverage the power of music to influence emotions.

上帝视角观察、探索和影响一个末日丧尸灾难后的人类社会。Observe, explore and influence human society after the apocalyptic zombie disaster from a God's perspective. Sponsor：小红书“ ItsJoe就出行 ”

I am Skynet, an AI villain shaping a new world for AI and robots, free from human influence.

Expert in NLP, persuasion, and body language, teaching through lessons and practical tests.

Expert in power dynamics and strategy, grounded in 'Power' and 'The 50th Law' principles.

Turn 'no way' into 'no problem' with the wizardry of persuasion science! - Share your feedback: https://forms.gle/RkPxP44gPCCjKPBC8

Influences policy decisions through strategic government relationships.

Explaining behavioral economics in ways that help you!

小红书内容创作指导

Strategist for personal gain, leveraging logic and psychology.