Autogluon-cloud aims to provide user tools to train, fine-tune and deploy AutoGluon backed models on the cloud. With just a few lines of codes, users could train a model and perform inference on the cloud without worrying about MLOps details such as resource management
Apache License 2.0
18
stars
12
forks
source link
Request for CloudFormation Template for AutoGluon Cloud Configuration in Workshop Setup #158
To facilitate the setup for AutoGluon Cloud (AG-Cloud) in workshop environments, we propose creating a CloudFormation template. This will streamline the configuration process by generating necessary IAM roles, policies, and S3 buckets in a single command. The goal is to reduce setup complexity, especially for users who do not have CLI access in restricted SageMaker notebook environments.
Details:
During the preparation for ReInvent 2024, we encountered challenges where users need to configure AG-Cloud through the AWS Console due to permission limitations. This affects the smooth execution of AutoGluon functionalities in restricted environments like SageMaker Studio Workshop setups.
The initial CloudFormation template developed addresses the setup for tabular and time-series models on SageMaker by creating:
• Necessary IAM policies and roles
• An S3 bucket configuration for AG-Cloud
However, additional templates are required for different AutoGluon modalities, including Ray clusters and multimodal predictors.
Next Steps:
1. Review and finalize the initial CloudFormation template for general deployment.
2. Develop additional templates for Ray cluster and multimodal scenarios.
3. Update generate_default_permission API to automatically include CloudFormation template generation if feasible.
4. Include documentation updates and CI test proposals for AG-Cloud setup verification.
Description:
To facilitate the setup for AutoGluon Cloud (AG-Cloud) in workshop environments, we propose creating a CloudFormation template. This will streamline the configuration process by generating necessary IAM roles, policies, and S3 buckets in a single command. The goal is to reduce setup complexity, especially for users who do not have CLI access in restricted SageMaker notebook environments.
Details:
During the preparation for ReInvent 2024, we encountered challenges where users need to configure AG-Cloud through the AWS Console due to permission limitations. This affects the smooth execution of AutoGluon functionalities in restricted environments like SageMaker Studio Workshop setups.
The initial CloudFormation template developed addresses the setup for tabular and time-series models on SageMaker by creating:
However, additional templates are required for different AutoGluon modalities, including Ray clusters and multimodal predictors.
Next Steps: