#ONNX

Lightning-Bolts extends PyTorch Lightning with additional components such as callbacks and datasets, facilitating applied research and production. It includes features like the Torch ORT Callback for enhanced training and inference speeds on NVIDIA or AMD GPUs. The SparseMLCallback further introduces sparsity in model fine-tuning using the DeepSparse engine. Encouraging community contributions, Lightning-Bolts evolves to support diverse machine learning needs in different domains without exaggeration.

Netron is a multi-device tool for viewing a broad range of machine learning models, supporting formats like ONNX and TensorFlow Lite, with additional experimental framework compatibility. Installable on major platforms, it simplifies the understanding of complex models for developers, researchers, and data scientists.

Tract is a versatile neural network inference engine supporting ONNX and NNEF model optimization and execution. It efficiently converts models from TensorFlow and Keras, making it suitable for both embedded systems and larger devices. Supporting models like Inception v3 and Snips, it runs efficiently on Raspberry Pi. As an open-source project under Apache/MIT licenses, it invites community contributions for custom application development.

A GPU-accelerated ONNX inference runtime entirely built in Rust, designed for web use. It supports Vulkan, Metal, and DX12, providing ease in model handling via CLI, Rust, Python, and WebGPU with WebAssembly. Available across platforms such as Windows, Linux, macOS, and Android. Includes comprehensive examples, CLI tools, and extensive documentation, catering to developers needing efficient, cross-platform inference solutions with Rust. Compatible with models like Squeezenet, MNIST, and BERT without overstated claims.

Optimize deep learning workflows with the TensorRT backend designed for ONNX model execution. This project aligns with TensorRT 10.5, ensuring full-dimension and dynamic shape processing. It integrates seamlessly with C++ and Python tools such as trtexec and polygraphy, enhancing model parsing efficiency. Comprehensive documentation, including FAQs and changelogs, aids in adaptive CUDA environment setups, making it a robust choice for ONNX deployment across experience levels.

The onnx-go project offers tools for Go developers to easily integrate ONNX models, enabling neural network capabilities without requiring advanced data-science skills. The library converts ONNX binary models into a format compatible with Go's ecosystem for effortless integration. Actively maintained by Orama, the project plans to expand support for ONNX operators, improving compatibility with models from the ONNX model zoo. It is suitable for developers looking for simple machine learning integration within Go applications.

Brocolli is a discontinued tool for converting PyTorch models to Caffe and ONNX using Torch FX. Although it is no longer maintained, it provides detailed instructions for model conversion. It supports model quantization and works well with popular models, such as AlexNet. A QQ group is available for community interaction and user support.

This project provides a complete C++ solution for Stable Diffusion image synthesis, eliminating the need for Python and enhancing deployment efficiency. It includes txt2img, img2img, inpainting functions, and ControlNet for guided generation. The library, optimized for DirectML, is aimed at real-time graphics and game developers, offering GPU-accelerated feature extraction. Prebuilt NuGet packages are available for seamless integration into Visual Studio C++ projects.

Fastembed-rs provides a Rust-based solution for generating embeddings with ONNX inference, supporting synchronous operations and parallel embeddings using Rayon without Tokio. It integrates @huggingface/tokenizers for fast text encoding and offers high-performing text and image embedding models, like Flag Embedding. The library is lightweight, with no hidden dependencies, and achieves high accuracy, outperforming models like OpenAI Ada-002. It's versatile, with support for custom models and local file inference.

Discover NVIDIA's TensorRT open-source components, including plugins and ONNX parser support. The repository provides sample apps showcasing platform capabilities, enhancements, and fixes. Developers will find coding guidelines and contribution instructions helpful. The Python package facilitates installation, compatible with CUDA, cuDNN, and vital tools for smooth deployment. Engage with the TensorRT community for updates and enterprise support through NVIDIA AI Enterprise. Access detailed developer guides and forums for further assistance.

NeoML is a versatile machine learning framework for developing, training, and deploying models on platforms including Windows, Linux, macOS, iOS, and Android. It is designed for applications such as computer vision, natural language processing, and OCR. NeoML supports various neural network layers and machine learning algorithms and executes tasks on both CPU and GPU. It integrates with Python, C++, Java, and Objective-C, supports ONNX format for compatibility, and offers features like multi-threading and GPU optimization, enabling efficient processing of both structured and unstructured data.

KotlinDL provides a high-level API inspired by Keras for deep learning, utilizing Kotlin. It integrates with TensorFlow Java and ONNX Runtime APIs, facilitating new model training and transfer learning with existing models. KotlinDL is designed for seamless deployment on JVM and Android, featuring simple configuration and comprehensive documentation with guides and tutorials. It enhances performance with GPU support and offers compatibility with Java logging frameworks.

Discover a repository that facilitates effortless inter-conversion of AI models among TensorFlow, PyTorch, ONNX, and other significant frameworks. With support for diverse quantization methods and optimization processes, this project enhances model performance across platforms like EdgeTPU and CoreML. It encourages community contributions for sample codes while keeping you informed on the progress in model conversion techniques for streamlined deployment.

onnx2torch allows for straightforward conversion of ONNX models to PyTorch, with support for adding custom layers and converting models back to ONNX. While support for current models and operations is limited, the tool encourages community feedback to enhance its coverage. It's easily integrable into existing workflows, supporting popular models like DeepLabV3+, YOLOv5, ViT, and GPT-J. Installable via pip or conda, it adapts to various opset versions, optimizing performance and flexibility.

The library simplifies the process of exporting, quantizing, and optimizing ESPnet models to the ONNX format independently of PyTorch. It facilitates ASR and TTS demonstrations on Google Colab utilizing pre-existing models without additional requirements. Capable of handling both pretrained and custom models, it provides detailed configuration options and supports GPU inference to enhance processing speeds. It also offers extensive installation and deployment guidelines to assist developers in integrating the library across multiple environments effectively.

Silero VAD is a pre-trained solution for voice activity detection, notable for its accuracy and speed. Supporting over 6000 languages and multiple sampling rates, it is adaptable to diverse audio environments. Lightweight and portable, it utilizes PyTorch and ONNX ecosystems for broad application in IoT, mobile, telephony, and voice interfaces. Free from restrictions, it ensures privacy with no telemetry or vendor lock-in, offering robust performance suitable for real-time detection across varied use cases.

ONNX is a community-driven, open-source project that provides a standardized model format to support interoperability across various AI frameworks and tools. It addresses the needs of both deep learning and traditional machine learning, facilitating smooth framework transitions and aiding swift research-to-production deployment. By engaging a wide range of frameworks, tools, and hardware, ONNX enhances AI research and development. The platform invites community collaboration to continually refine its offering, fostering ongoing improvements to support the dynamic field of AI innovation efficiently.

Explore high-quality speech-to-text (STT) and text-to-speech (TTS) models designed for simplicity and performance. These models enable seamless, natural-sounding speech conversion across multiple languages, including Russian, English, German, and Spanish. Enhance text readability with automatic punctuation and capitalization, all with minimal setup using PyTorch, pip, or manual caching. Achieve efficient and reliable outcomes suited for diverse speech and text processing applications.

tensorflow-onnx efficiently converts models from TensorFlow, Keras, Tensorflow.js, and Tflite to ONNX, supporting multiple TensorFlow versions, Python environments, and ONNX opsets. It provides developers with a reliable solution for seamless model migration, enhancing cross-platform compatibility and performance. Features include detailed CLI usage, troubleshooting documentation, and installation guides. Regular updates ensure support for emerging frameworks and experimental features, aligning with current AI workflows.

The project offers a detailed guide for leveraging ONNX and YOLOv8 in object detection, covering essential requirements and installation procedures, with a focus on GPU compatibility using onnxruntime-gpu for NVIDIA devices. It facilitates model conversion via Google Colab using clear Python scripts. The repository features examples for image, webcam, and video inference, demonstrating the model's versatility and efficiency. This project serves as a valuable resource for developers interested in implementing YOLOv8 in ONNX format, providing practical application insights.

The onnx2tflite project simplifies the conversion of ONNX models to TensorFlow Lite, maintaining high precision and efficiency. It minimizes errors per element to less than 1e-5 compared to ONNX outputs and accelerates model output, achieving speeds 30% faster than other methods. This tool automatically aligns channel formats and supports the deployment of quantized models, including fp16 and uint8. Users can modify input and output layers, add custom operators, and benefit from a straightforward code structure, making it a versatile solution for AI model deployments.

Paddle2ONNX facilitates the transformation of PaddlePaddle models into the ONNX format, supporting integration with multiple inference engines such as TensorRT and OpenVINO. Compatible with PaddlePaddle version 2.6.0 and ONNXRuntime 1.10.0 and above, it offers command-line tools for model parameter adjustments and ONNX model optimization, enhancing deployment flexibility across platforms. Developers can also engage in its development by following contribution guidelines, making it a valuable tool for cross-platform machine learning deployment.

YOLOv8-TensorRT boosts YOLOv8 performance by employing TensorRT for faster inference. It leverages CUDA and C++ for engine construction and facilitates ONNX model export with NMS integration. This project provides flexible deployment options using Python and Trtexec on various platforms, including Jetson. The comprehensive setup guide helps adapt to different AI deployment needs, offering an efficient PyTorch alternative.

Microsoft Cognitive Toolkit (CNTK), an open-source deep learning framework, models neural networks as computational graphs for seamless execution of architectures like DNNs, CNNs, and RNNs. It incorporates stochastic gradient descent with automatic differentiation, supporting multi-GPU and server parallelization, suitable for intensive deep learning applications. Though major updates have halted, CNTK maintains compatibility with the latest ONNX standards, promoting AI framework interoperability. Extensive resources are available for users to explore and optimize the toolkit’s features.

Explore a framework designed to streamline the creation of LLM applications tailored for business use. By utilizing compact, specialized models, it ensures secure and cost-effective integration with enterprise resources. The framework includes a comprehensive RAG Pipeline to connect AI models efficiently and offers over 50 specialized models aimed at automating enterprise tasks. This solution facilitates development without demanding high-end hardware, allowing application development on standard laptops.

ONNXMLTools enables the conversion of machine learning models from toolkits like TensorFlow, scikit-learn, and Core ML into the ONNX format, supporting frameworks such as LightGBM, XGBoost, and CatBoost. This tool improves model interoperability and deployment across platforms. Installation via PyPi or source builds are available, with Python 3.7+ compatibility and specific framework dependencies. Detailed examples and extensive testing bolster reliability in model conversions.

OOTDiffusion employs outfitting fusion via latent diffusion technology to deliver controllable virtual try-ons, with model checkpoints from VITON-HD and Dress Code datasets supporting both half and full-body garments. Now ONNX compatible, OOTDiffusion offers human parsing solutions for modeling on powerful GPUs, tested on Ubuntu 22.04. Installation requires setting up a conda environment and the necessary packages.

Discover a user-friendly NVIDIA TensorRT wrapper for deploying ONNX models in C++ and Python. Despite its lack of ongoing maintenance, tiny-tensorrt emphasizes efficient deployment using minimal coding. Dependencies include CUDA, CUDNN, and TensorRT, easily setup through NVIDIA's Docker. With support for multiple CUDA and TensorRT versions, it integrates smoothly into projects. Documentation and installation guidance are available on its GitHub wiki.

Addressing key transpose challenges, this tool converts ONNX NCHW models to TensorFlow, TFLite, or Keras NHWC formats with ease. It is regularly updated to improve model optimization and bug resolution, encouraging community participation through pull requests. If issues occur, using older versions is advised. The tool also suggests transitioning to torch.onnx.dynamo_export for better conversion capabilities, with its future tied to runtime advancements. Discover its role in model conversion for enhanced efficiency.

This project allows scikit-learn models to be transformed into ONNX format for optimized performance with ONNX Runtime. Compatibility with opset 21 and support for a variety of external converters extend its functionality across numerous models. Easily installable via PyPi or source, the project offers extensive documentation and community support for smooth integration. Contributions are warmly accepted under the Apache License v2.0.