|
Work item:
|
F.AICP-MDep
|
|
Subject/title:
|
Technical specification for artificial intelligence cloud platform: AI model deployment
|
|
Status:
|
Under study
|
|
Approval process:
|
AAP
|
|
Type of work item:
|
Recommendation
|
|
Version:
|
New
|
|
Equivalent number:
|
-
|
|
Timing:
|
2025 (Medium priority)
|
|
Liaison:
|
ITU-T SG13, JCA-ML, ISO/IEC JTC1 SC42
|
|
Supporting members:
|
Zhejiang Lab, MIIT, Zhejiang Dahua Technology, Alibaba.
|
|
Summary:
|
Model deployment is the process of integrating the trained AI models that have completed the development process into production to make decisions on real-world data. With the wide application of artificial intelligence, it has become one of the most important stage in the machine learning life cycle. However, Model deployment can be a quite hard and time-consuming process. How to deploy the AI models quickly with the lowest deterministic latency on a real-time performance platform is a big challenge for developers and users.
AI cloud platform supports the end-to-end machine learning lifecycle from designing to production: data exploring, model development, model deployment and continuous monitoring. It simplifies and automates the process of AI model deployment to enable AI engineers even without prior knowledge in AI achieve the deployment of trained AI models. More and more platforms are developed to help AI engineers deploy their AI models quickly and flexibly. However, the incomplete functionality, complicated structures and usage differences of AI cloud platforms lead to a steep learning curve for developers and users.
This draft Recommendation aims to standardize the model deployment process and specify the function requirements of AI cloud platform to enable the process of AI model deployment. On the one hand, it can provide guidance for AI engineers to deploy models and develop AI cloud platform; on the other hand, it can be a supplement to existing standards to complete the full life cycle management of AI models.
|
|
Comment:
|
-
|
|
Reference(s):
|
|
|
Historic references:
|
|
Contact(s):
|
|
| ITU-T A.5 justification(s): |
|
|
|
|
First registration in the WP:
2023-08-21 15:56:08
|
|
Last update:
2023-12-18 11:10:45
|
