Author: Borja Gomez (franciscodebo@gmail.com)
MLOps-3
Barcelona, May 2022
This is the Mini-Capstone assigment for the cohort #3 of the MLOps program.
- Access to production: http://34.139.170.162/
All the commends are implemented using Makefile. Here you have the list of available commands:
- Build the image. Do this any time you commit code or add a new library:
make build
- Run the container associated to the image. The web server is running so every change at any file it will be refreshed. Don't run locally. Run via de Docker container!
make run
- Stop the app
make stop
- Access the container in the console
make bash
- Access the logs
make logs
- Push the image to the public ECR at GCP and do a rolling update of the Kubernetes deployment with the new image.
make deploy
- Authenticate gcloud and docker
make auth
- Create the infrastructure. Just need it to run once!
make create-infrastructure
This project is based on the following references:
- Resnet Image Classification Webapp (Abhishek Nagaraja - https://github.com/anagar20/Resnet-Image-Classification-Flask-App).
- CIFAR 100 transfer learning with ResNet50: https://www.kaggle.com/code/saileshnair/cifar-100-transfer-learning-resnet50/notebook