Sakana Ai’s New Agent Framework Can Improve Model Deployment Speed

Sakana Ai, A Tokyo-Based Artificial Intelligence (AI) Firm, introduced a new artificial intelligence (AI) Agentic Framework that can improve the development and deployment speeds of large language models (llms). Announced on Thursday, the company unveiled the ai cuda engineer that improves bot the pre-training and infection speeds of an ai model by optimising the codbaase. The AI ​​firm highlighted that the entry process is driven by ai agents and is end-to-end automated. Notable, Sakana Ai introduced The AI ​​Scientist Last Year which can Conduct Scientific Research.

Sakana Ai Unveils Ai Cuda Engineer

In a postThe japanese ai firm stated that after development ai systems that can create new models, and full automate the ai research process, it began working on Ways to Speed ​​Up The Deplorement and Infections of an anne Llm.

The company said that the research is the development of the ai cuda engineer. It is a fullly automated, Comprehensive Agent Framework for Cuda (Compute Unified Device Architecture) Kernel Discovery and Optimization.

Cuda kernels can be undersrstood as specialized functions that run on nvidia gpus, allowing parallel execution of code across multiple threads. Due to parallelism, it is more optimized than traditional methods and allows for the acceleration of computational tasks, especially that with large datasets. As such, this is considered a great way to optimise ai models’ deployment and infection.

Sakana ai said the Ai Cuda Engineer Can Automatically Pytorch Modules Into Optimized Cuda Kernels, to significantly improve the deprive deflioment speedups. It can generate kernels that are said to be 10-100 times faster than its pytorch counterpart.

The process includes four steps. First, the agent framework converts the pytorch code into working kernels. Then, the agent implements optimization technique to ensure only the best kernels are generated. Then, Kernel Crossover Prompts are added, which combine multiple optimized kernels to create new kernels. Finally, the AI ​​agent preservs the high-PeerforMance Cuda Kernels in an archive, which are used to delivery performance improvements. The company has also published a study That further details the process.

AlongSide the Paper, Sakana AI is also Publishing the Ai Cuda Engineer Archive, which is a dataset consisting of more than 30,000 kernels generated by the Ai. These kernels are released under the cc-by-4.0 license and can be accessed via hugging face.

Additional, The Japanese firm also launched a website that lets visitors interactively explore 17,000 verified kernels and their profiles. The website allows users to explore these kernels 230 tasks, and also lets them compare cuda kernels across individual experiences.