We use DIS segmentation model wrapped in a service with option to deploy in sagemaker Below is the link to the model https://github.com/xuebinqin/DIS
The repo contains
- Option to start the service locally
- Deploy the model in Sagemaker
- CI/CD pipeline to trigger tests and deploy on pull request comment.
Starting the service locally
- Please put the saved model in src/saved_model before building docker image
docker build -t dis-segmenter .
docker run -p 8080:8080 --rm dis-segmenter serve
- The classifier should have test for all the method including loading and predictions
- Will implement api level testing to test the input and output formats of endpoints
- Would do the model loading on application startup and not on first request
- will implement the ping health check for sagemaker endpoints
- Use typing in all places and do mypy check
- Use object models (pydantic, cattr) for request and response data models.