Sentimental analyses with MLFLOW and models Wrappers

Table of content

• Overview
• Architecture
• Install
• Usage
• Contributing
• Production
• Monitoring
• Api
• License
• Author
• Thanks

Overview

Tweet sentimental analyses with different models.

Four wrapper of models:

• Logistic Regression
• Random Forest
• LightGBM
• Bert
• Roberta
• LSTM

MLFlow is used to list all experiments and easily commpare results for several differents configurations and select the bests

Optuna is used to optimise parameters. It run a set of experiments with a variation of parameters and select the best configuration maximising the accuracy.

Architecture

The application contains alerting system and monitoring on grafana on port 3000 APP PORT MLFLOW 5001 API 5000 GRAFANA 3000 MONGO PROMETHEUS LOKI

A loki message and prometheus services are define Loki message show the new tweets on grafana. New tweets are saved on a Mongo db. Prometheus send metrics as the number of prediction running. An alert is send by mail when number of predictions in concurrency are up to 5. An alert is send when the result of the prediction is too bad, probability < 0.5.

Install

The app is dockerised and can be installed launching the command

docker compose up 

or to run in background

docker compose up -d 

Contributing

Install uv (Rust package to fastly install package)

curl -Ls https://astral.sh/uv/install.sh | bash
export PATH="$HOME/.cargo/bin:$PATH"

Source the code in the container

Modify the docker-compose.yaml to add the source code as volume

volumes:
  - ./src:/app/src/
  - ./mlruns:/app/mlruns/

Usage

OVH Train with AI train

Create an object storage on OVH managed with ovhai cli The secret key is obtain clicking on the user object storage line 'access secret key'

ovhai datastore add s3 datastore-model https://s3.gra.io.cloud.ovh.net/ gra <acces_key> <secret_key> --store-credentials-locally
# Upload to training file
ovhai bucket object upload datastore-model@GRA ../data/training.1600000.processed.noemoticon.csv  --object-name training.1600000.processed.noemoticon.csv

1 Datastore is associate to one bucket, it is a gateway Credentials are stored in ~/.config/ovhai/context.json

uv pip install boto3 awscli ovhai

Run on multi GPU

DEBUG

export TORCH_DISTRIBUTED_DEBUG=DETAIL

python -m torch.distributed.run --nproc_per_node=2 train.py

Tests

pytest src/tests

Launch a test to verify the prection from the API

Go on 127.0.0.1:5000, tap your tweet and click on predict button

Production

An exemple deployment is available on https://tweetsentiment.shift.python.software.fr

Monitoring

Add alert and monitoring and dashboard on grafana on your local instance and save them in grafana folder. Reload grafana and they will be available on http://localhost:3000 as provisionning templates

Api

You can contact the api example or change the url on the script predict_client.py to test your instance

export $(cat .env | xargs)
python predict_client.py 

License

MIT License

Author

Shift python software

Thanks

Thanks to all contributors