MLOPS PROJECT

Idea of this project is to have simple cat vs dog classification project with emphasize on MLOps pipeline. Goal is to have everything running on AWS.

MLFlow for model versioning

1. pip install mlflow
2. mlflow ui
3. Go to http://localhost:5000

LakeFS for data versioning

1. Follow these steps in order to set up lakeFS https://towardsdatascience.com/data-versioning-all-you-need-to-know-7077aa5ed6d1

Running pipeline

1. pip install requirements
2. Create config.ini with your credentials. Example is in config.ini.EXAMPLE
3. Prepare data and put it under dataset_cat_vs_dog folder to have sub folders dog and cat with images.
4. python train.py
5. export PYTHONPATH=$PYTHONPATH:$pwd
6. python prepare_and_export/export.py
7. docker run -t --rm -p 8501:8501 -v "PATH_TO_PROJECT/MLOps/serving:/models/dog_detector" -e MODEL_NAME=dog_detector tensorflow/serving &
8. python prepare_and_export/inference.py