Unfortunately I did not have enough time to complete this project within the cohort period. I will finish it later little by little to improve my knowledge in MLOps.
Repository for training MLOps for 2023 cohort from DataTalks.club. An end-to-end ML pipeline for NLP processing at B2W dataset.
Table of sections:
With the use of Machine Learning Operations (MLOps), this project aims to:
- Predict the sentiments of customers
- Model topics to understand customer review content.
Specifically for me, I combined my theoretical knowledge on NLP with MLOps to increase my challenge with MLOps, while understading NLP more deeply, mainly for project architecture design.
This project is structured as follows:
- monitoring is a folder to organize services responsible for track some aspect of the project (e.g mlflow for experiments tracking and prefect for pipeline orchestration)
- training is a folder to models development for sentimental analysis and topic modelling.
Before execute this project, be sure to attend the requisites:
- docker. You can use this command below for some linux distributions.
curl -fsSL https://get.docker.com | bash- Python 3.10. You can install through anaconda, venv or directly build from source.
To start the services to monitor any modifications, run:
docker compose --env-file ./monitoring/mlflow/.env up -d --build