Preferred Networks is a Japanese company that has been making waves in the field of artificial intelligence. They're known for their research and development of neural networks.
Their work has led to the creation of the CHAINER library, which is a popular tool for deep learning. This library has been widely adopted by researchers and developers around the world.
One of the key features of CHAINER is its ability to handle large-scale neural networks. This makes it an ideal choice for complex tasks such as image recognition and natural language processing.
Challenge
Preferred Networks, a Japanese company, faces a significant challenge in delivering its AI/ML resources to researchers in a user-friendly, fair, and flexible manner.
The company operates clusters across more than three locations, providing the necessary computational resources for its researchers.
These clusters are heterogeneous, comprising multiple series of GPUs and MN-Cores, which the company has developed and researched.
The team managing this infrastructure had only eight engineers at the time, making operations automation an essential focus.
The company has over 1500 GPUs and more than 200 MN-Cores, which adds to the complexity of managing these resources efficiently.
Suggestion: California Resources Corporation
Impact and Benefits
Preferred Networks achieved significant efficiency improvements and scalability by adopting Kubernetes, making their AI/ML platform more user-friendly.
By developing a device plugin for MN-Core, it became a first-class resource like CPU, memory, and GPU, accessible to researchers.
Using Kustomize to standardize Kubernetes manifests across various sites, the company can add new sites within three months with three engineers.
A node can be recreated within 30 minutes, from OS installation to provisioning and joining the Kubernetes cluster, making it easy for administrators to reset nodes.
The flexibility of the Kubernetes scheduler was utilized to develop an extended scheduler, allowing the specification of resource allocation for researchers, including the number of accelerators.
By creating a scheduler plugin for Kubernetes, Preferred Networks made unique enhancements for distributed learning, such as the “gang scheduling” method, enabling more efficient use of cluster resources.
This scheduler is open source.
By extending Kubernetes features, Preferred Networks added a custom specification syntax for resources, allowing researchers to specify particular GPUs, for example, preferred.jp/gpu-v100-24gb: 1.
Applications and Industries
Preferred Networks is a company that has made significant contributions to the field of deep learning, particularly in the area of computer vision. They have developed the Chainer framework, a popular open-source deep learning library.
Their technology has been applied in various industries, including robotics and manufacturing. In these fields, Preferred Networks' deep learning solutions have improved efficiency and accuracy.
One notable example is their work with the Toyota Motor Corporation, where they developed a deep learning-based system for inspecting automotive parts. This system was able to detect defects more accurately and efficiently than traditional methods.
Expand your knowledge: Broadcasting Satellite System Corporation
K8s for User-Friendly AI/ML Platform
Kubernetes (K8s) is a game-changer for deploying and managing AI/ML models.
It allows for efficient resource allocation, ensuring that models are running smoothly and without interruptions.
K8s provides a scalable and flexible platform for AI/ML workloads, making it easier to handle large datasets and complex computations.
By automating deployment, scaling, and management, K8s reduces the administrative burden on developers and data scientists.
This means they can focus on what matters most - building and refining AI/ML models that drive business value.
K8s also enables seamless integration with popular AI/ML frameworks and tools, such as TensorFlow, PyTorch, and scikit-learn.
This integration simplifies the development process and accelerates time-to-market for AI/ML applications.
By providing a user-friendly interface and robust features, K8s empowers developers and data scientists to build and deploy AI/ML models with ease.
Robotics
Robotics is a rapidly evolving field with many exciting applications.
Preferred Robotics is a subsidiary of Preferred Networks that was established on November 1, 2021. It raised 2 billion yen from Amano Corporation shortly after its establishment.
In March 2022, Preferred Robotics secured additional funding from Asahi Kasei and Sumitomo Mitsui Banking Corporation, raising a total of 600 million yen.
The company co-developed the autonomous small floor cleaning robot HAPiiBOT (HapiiBot) with Amano Corporation in September 2022.
Frequently Asked Questions
Is Preferred network a public company?
No, Preferred Networks is a privately held company, meaning it is not publicly traded on any stock exchange. You can find information on its past funding rounds and stock prices through filings and secondary trading data providers.
Who is the CEO of Preferred network?
The CEO of Preferred Networks is Toru Nishikawa, who is also the co-founder of the company. He leads the organization as its Chief Executive Officer.
Featured Images: pexels.com
