We are excited to bring Transform 2022 back in person from July 19th, effectively July 20th to 28th. Join an AI and data leader for insightful talk and exciting networking opportunities. Register today!
Performing the life cycle of a complete machine learning workflow can often be a complex operation involving multiple disconnected components.
Users must have machine learning-optimized hardware, the ability to adjust workloads across that hardware, and some form of machine learning operation (MLops) technology to manage their models. Artificial intelligence (AI) computing orchestration vendor Run: ai (raised $ 75 million in March) and MLops platform vendor Weights & Biases (W & B) have partnered with Nvidia to ease the burden on data scientists. ..
“This three-party partnership allows data scientists to plan and run models using Weights & Biases,” Omri Geller, CEO and co-founder of Run: AI, told VentureBeat. “In addition, Run: ai tunes all workloads in an efficient way on Nvidia’s GPU resources, giving you a complete solution from hardware to data scientists. . “
Run: ai is designed to allow organizations to use Nvidia hardware for machine learning workloads in cloud-native environments. This is a deployment approach that uses containers and microservices managed by the Kubernetes container orchestration platform.
One of the most common ways organizations can perform machine learning in Kubernetes is to use a Kubeflow open source project. Run: ai is integrated with Kubeflow to help users optimize their use of Nvidia GPUs for machine learning.
Omri added that Run: ai is designed as a Kubernetes plugin that enables virtualization of Nvidia GPU resources. By virtualizing the GPU, you can divide resources so that multiple containers can access the same GPU. You can also use Run: ai to manage quotas for virtual GPU instances so that workloads always have access to the resources they need.
According to Geller, the goal of this partnership is to make the complete machine learning operation workflow easier for enterprise users to use. To that end, Run: ai and Weights & Biases are building an integration to make it easier to run the two technologies together. Prior to the partnership, Omri said organizations that wanted to use Run: ai and Weights & Biases had to go through a manual process to bring the two technologies together.
Seann Gardiner, Vice President of Business Development at Weights & Biases, said the partnership will allow users to take advantage of the training automation provided by Weights & Biases using GPU resources tuned by Run: ai. I commented.
Nvidia isn’t monogamous, it’s partnering with everyone
Nvidia is partnering with both Run: ai and Weights & Biases as part of its larger strategy of partnering within the machine learning ecosystem of vendors and technology.
“Our strategy is to partner fairly and evenly with the comprehensive goal of ensuring that AI is ubiquitous,” Scott McClellan, Senior Director of Product Management at Nvidia, told Venture Beat. ..
McClellan said the partnership between Run: ai and Weights & Biases was particularly interesting. In his view, the two vendors offer complementary technologies. Both vendors can now plug into the Nvidia AI Enterprise platform, which provides software and tools to make AI available to the enterprise.
McClellan said the three vendors are working together and if data scientists are going to use Nvidia’s AI enterprise containers, they don’t need to understand how to do their own orchestration deployment framework or their own scheduling. rice field.
“These two partners complete our stack, or we complete their stack and complete each other’s stack, so the whole is greater than the sum of the parts.” He said.
Avoid the “Bermuda Triangle” in MLOps
For Nvidia, partnerships with vendors such as Run: ai and Weights & Biases are helping many companies solve the key challenges they face when they first embark on an AI project.
“When a data science or AI project is about to move from experimentation to production, this can be a bit like the Bermuda Triangle, where many projects die,” McClellan said. “That is, it disappears in the Bermuda Triangle. How can I move it to production?”
McClellan hopes that Kubernetes and cloud-native technologies commonly used by today’s enterprises will make it easier than ever to develop and operate machine learning workflows.
“MLops is the developer of ML. Literally, they don’t die when they go into production and they can lead a fulfilling and healthy life,” says McClellan.