From ML experiments to production:

Versioning and Reproducibility
with MLV-tools

Stéphanie Bracaloni and Sarah Diot-Girard

About Us

Sarah Diot-Girard
Data Scientist since 2012
Interested in DataOps and Ethics
PeopleDoc logo @SgdJlbl
Stéphanie Bracaloni
Software Engineer since 2013
Automation and Code Quality
PeopleDoc logo @sbracaloni
PeopleDoc logo

The story, all names, characters, and incidents portrayed in this production are fictional.
All similarities with existing past or future Data Science projects are purely coincidental.

Monday morning, 9am

Confusion Matrix

A confusion matrix showing nice ML results

ROC Curve

A nice ROC curve
Monday morning, 10am
Two weeks later


vs Data Scientist

vs Software Engineer

Two months later

Confusion Matrix

A confusion matrix showing nice ML results

ROC Curve

A nice ROC curve

├── raw_data.csv
├── cleaned_data.csv
├── cleaned_data_final.csv
├── cleaned_data_preprocessed_final.csv
└── cleaned_data_preprocessed_final2.csv

Data Versioning


What's not to love?

  • Great for code
  • Collaborative
  • Everyone knows (of) it

Huge datafiles

  • Bloats the repository
  • Binary?
  • So slow and tedious...

Git LFS / Git annex

What's not to love?

  • Great for handling larger files
  • Git integration
  • Open-source
Data + code + hyperparameters = metrics

├── raw_data.csv		mar   8		12:17
└── preprocessed_data.csv 	feb   26	13:24
Data x code x hyperparameters = metrics

Is DataOps solved?

Not quite yet

For the data scientists

  • Modular improvements
  • Fast iterations
  • Confidence in results
  • Cannot loose data

For the engineers

  • Full power of IDE
  • Functional tests
  • CI and production-ready

Current limitations

  • DVC is another CLI tool
  • Git and DVC checkout for switching branches

Under development

  • Checking if all scripts are up-to-date when pushing
  • Easily comparing metrics between experiments
  • Pipeline packaging (with Debian and Anaconda)
  • Dynamically handling of hyperparameters

Future work

DVC pipelines as powerful as scikit-learn pipelines
  • Cross-validation
  • Hyperparameter tuning

Future work

We want to hear about your use cases!

We are waiting for your Pull Requests!

PeopleDoc logo
PeopleDoc logo

PeopleDoc logo

Fonts: White Rabbit by Matthew Welch
Capsuula by Henrich Fichna
Icons made by Smashicons, Good Ware, Designmodo, Freepik from; license CC 3.0 BY

Contact Us !