zhreshold / gluoncv-sagemaker-serving

SageMaker serving for GluonCV models

gluoncv-sagemaker

SageMaker serving for GluonCV models

Pre-requisite

• SageMaker notebook(recommended)
• Computer with AWS CLI configured.

Usage:

The following instructions works for SageMaker notebook instance(make sure ECR write permission is added in SageMaker IAM role)

• First activate a mxnet_p36 env by source activate mxnet_p36 in a new terminal.
• Then install required python packages: pip install -r requirements.txt
• Execute the batch building script: python batch_build.py.

The script will build ModelPackage for models listed in image_classification.txt and object_detection.txt. Due to the validation process, it might take very long time to finish.

Marketplace listing Metadata Generator

python batch_describe.py

The sample output looks like: https://gist.github.com/zhreshold/f616f5b894d386701b4c85a4b40d200c

TODO

• Automatic ARN injection for built ModelPackage
• Object Detection metadata
• Semantic Segmentation metadata