filmov
tv
Scaling ML Workloads on Amazon EC2 | Performance or Productivity in AI | Intel Innovation 2022
Показать описание
In this video, Sundar Ranganathan, Head of ML Frameworks Specialists, AWS, talks about the challenges with the exponential growth of SOTA AI models, strategies for distributed training and inference, a customer example of a large scale AI and HPC implementations.
To move AI from research/innovation to mainstream, there are, even today, significant barriers that exist to both build and deploy models to scale. To develop and deploy models, a data scientist/AI developer is doing all kinds of tasks that are tangential to data science but less actual data science. The chart you’re looking at reflects research we did at our own Mlcon machine learning conference and surveys of our own customers. All of these incidental tasks are necessary at some level, but they all add friction to the process, delay time to value, and contribute to lost opportunities.
So what are some of these incidental tasks? Installing and configuring hardware like GPUs, CPUs, accelerators, and storage, standing up cloud computing resources, configuring containers and container orchestration software like Kubernetes, figuring out how to manage hybrid cloud infrastructures. All of this reduces the ROI of AI from a business perspective. There are tools that help reduce these complexities, but few have extensive MLOps capabilities, and they often add complexity with heavy code requirements or require expertise in tools like Kubernetes or Docker to take advantage of containerization for model standardization, management, and repeatability.
About Intel Software:
The Intel® Developer Zone encourages and supports software developers that are developing applications for Intel hardware and software products. The Intel Software YouTube channel is a place to learn tips and tricks, get the latest news, watch product demos from both Intel, and our many partners across multiple fields. You'll find videos covering the topics listed below, and to learn more, you can follow the links provided!
Connect with Intel Software:
#intelsoftware #developertools #ai #intelinnovation
Scaling ML Workloads on Amazon EC2 | Performance or Productivity in AI | Intel Innovation 2022
To move AI from research/innovation to mainstream, there are, even today, significant barriers that exist to both build and deploy models to scale. To develop and deploy models, a data scientist/AI developer is doing all kinds of tasks that are tangential to data science but less actual data science. The chart you’re looking at reflects research we did at our own Mlcon machine learning conference and surveys of our own customers. All of these incidental tasks are necessary at some level, but they all add friction to the process, delay time to value, and contribute to lost opportunities.
So what are some of these incidental tasks? Installing and configuring hardware like GPUs, CPUs, accelerators, and storage, standing up cloud computing resources, configuring containers and container orchestration software like Kubernetes, figuring out how to manage hybrid cloud infrastructures. All of this reduces the ROI of AI from a business perspective. There are tools that help reduce these complexities, but few have extensive MLOps capabilities, and they often add complexity with heavy code requirements or require expertise in tools like Kubernetes or Docker to take advantage of containerization for model standardization, management, and repeatability.
About Intel Software:
The Intel® Developer Zone encourages and supports software developers that are developing applications for Intel hardware and software products. The Intel Software YouTube channel is a place to learn tips and tricks, get the latest news, watch product demos from both Intel, and our many partners across multiple fields. You'll find videos covering the topics listed below, and to learn more, you can follow the links provided!
Connect with Intel Software:
#intelsoftware #developertools #ai #intelinnovation
Scaling ML Workloads on Amazon EC2 | Performance or Productivity in AI | Intel Innovation 2022
0:20:18
0:43:52
0:12:50
0:27:45
0:24:16
1:37:16
0:25:19
0:51:43
0:27:17
0:16:16
0:31:56
0:37:52
0:27:51
0:51:11
0:55:56
0:59:37
0:54:42
0:37:03
0:34:48
0:30:13
0:50:22
0:25:29
0:31:49
0:00:50