Skip to content

Nyun Zero

Overview

Nyun Zero is an efficiency-focused deep learning platform designed to streamline the entire development-to-deployment cycle of deep learning models. Zero offers multiple plugins covering common use cases in deep learning production cycles:

  1. ADAPT: A robust adaptation module that effortlessly fine-tunes and performs transfer learning on a diverse range of deep learning tasks and models. With ADAPT, users can adapt large and even medium-sized vision and language models seamlessly, enabling the creation of custom models with just a few clicks. This module incorporates state-of-the-art adaptation methods such as (Q,G)-LoRA, SSF, etc., allowing users to strike the optimal balance across various model metrics, including parameter count and training throughput.

  2. KOMPRESS: Developed to compress and optimize deep learning models, Kompress provides a set of compression techniques tailored for specific deployment constraints. Users have the flexibility to choose and combine multiple techniques to achieve the best trade-off between model performance and deployment constraints. Leveraging cutting-edge techniques like pruning, quantization, distillation, etc., Kompress achieves exceptional model compression levels on a variety of language and vision models.

  3. TRACK: Helps users track crucial model deployment metrics often neglected in the model building stage. These metrics include inference latency, carbon emission, energy consumption, etc. Track seamlessly works over multiple frameworks and models with support for even custom models.

Setup and Installation

Zero can be deployed on-premises, ensuring complete data privacy. Users can access the Zero Web App at zero.nyunai.com to log in and set up their infrastructure.

Additionally, NyunAI provides a centralized Docker Compose file to deploy the Zero SDK on-premise. Users can find the necessary resources at zero-docker. With just a few simple commands, users can set up all required dockers and packages to get started.

About NyunAI

NyunAI began its journey in 2020 with state-of-the-art research on model compression. Recognizing the challenges in setting up pipelines for model compression and downstream task adaptation, NyunAI aimed to simplify the process for researchers and developers, allowing them to efficiently build and deploy models while focusing entirely on problem-solving. Since then, NyunAI has remained committed to developing efficient deep learning technology and supporting software.

Contact NyunAI

NyunAI's support team is always available to assist users with any questions or concerns. Users can reach out to the team via email at support@nyunai.com.