Get MNN for Windows – Secure Program for Your PC

Download MNN

Screen Prints

Get MNN

MNN is selected extremely efficient and lightweight deep learning framework. It supports inference not to mention practice of layered learning designs, and has industry leading performance for inference and training on-device. At nearby, MNN possesses been integrated in more than 20 apps of Alibaba Inc, such as Taobao, Tmall, Youku, Dingtalk, Xianyu and etc., covering more than 70 usage scenarios such as live broadcast, short video capture, search recommendation, product searching by image, interactive marketing, justness distribution, security risk control. In addition, MNN is also used on embedded devices, such as IoT. MNN Workbench could be downloaded from MNN's homepage, for which provides pretrained models, visualized training tools, and one-click deployment with respect to models to devices. Artificial human foundation, core so size is about 400KB, OpenCL because is about 400KB, Vulkan because of this is regarding 400KB. Supports hybrid computing through multiple hardware. Currently supports ELECTRONIC BRAIN and GPU.

Features

• Implements core computing with heaps of optimized assembly code to make full use in the case of the ARM CPU
• For iOS, GPU acceleration (Metal) can be turned on, at which is expedited than Apple's native CoreML
• For Android, OpenCL, Vulkan, in conjunction with OpenGL are available and substantial tuned for widely accepted GPUs (Adreno and Mali)
• Convolution and transposition convolution algorithms are quick and permanent. Some other Winograd convolution algorithm is widely used to better harmonious convolutions such as 3x3 -> 7x7
• Twice speed increase for our new architecture ARM v8.2 with FP16 half-precision calculation support
• Optimized for devices, no dependencies, can be easily deployed to mobile contraptions and a variety of embedded devices

Project Blueprints Project Activity

See All Activity >

CategoriesFrameworks, Machine Studying, Deep Familiarity Frameworks, LLM Inference LicenseApache License V2.0 Follow MNN

MNN Web Site

Other Useful Business Product Our Free Tactics just got more desirable! | Auth0 by Okta

With up to 25k MAUs and unlimited Okta associates, our Free Plan lets you focus on what the participant do best—building great apps.

You demanded, we executed! Auth0 is excited to unfold our Free and Paid plans to include more options so you can focus on building, executing, furthermore scaling apps without obtaining to worry about your security. Auth0 currently, thank yourself later. Try free as of this moment Tariff This ProjectLogin To Rate This Project

User Reviews

Function the first to post every review of MNN! Additional Project Details Operating SystemsLinux, Android, Apple iPhone Programming LanguageC++ Related Categories C++ Frameworks, C++ Appliance Familiarity Software, C++ Deep Learning Frameworks, C++ LLM Inference Object Registered 2021-09-10 Similar Business Module

• FIRM Watson Machine Learning Accelerator Accelerate your serious comprehension workload. Zest the student's time among value in the company of MACHINE COGNITION model training what’s more inference. With advancements in interpret, framework and data access, conglomerates are adopting deep learning more widely to extract and scope insight through speech reputation, natural language processing... See App
• DeepCube DeepCube focuses on the research and development of deep skill-building frameworks that issue in improved real-world deployment of ALGORITHMIC INTELLIGENCE systems. The company’s plethora of innovative innovations include methods instead of faster and more faultless training of deep learning models and drastically improved inference... See Software
• NVIDIA Triton Inference Server SUPERMICRO Triton™ inference server fulfills fast and scalable AI in production. Open-source inference serving software, Triton inference gateway boosts AI inference via enabling clubs deploy trained AI models from any apparatus (TensorFlow, NVIDIA TensorRT®, PyTorch, ONNX, XGBoost, Python,... Stare System

Report unacceptable content