| Time: | April 27, 2022, 2:00 p.m. (CEST) |
|---|---|
| Lecturer: | F. Zills, J. Range |
| Download as iCal: |
|
The ML committee of the Cluster of Excellence SimTech hereby invites you to the upcoming ML Session with F. Zills and J. Range on Continuous Model Design and Simulation with CML and reporting on Workflows
Date: Wednesday, 27 April 2022
Time: 2 to 3 pm
Place: Pfaffenwaldring 57, SR 8.122 (in person, NOT by Webex)
The design of machine learning models involves many crucial steps, which can be broken down to model discovery followed by hyperparameter-tuning and finally the deployment for production. Currently, most of these steps are executed and evaluated manually, which is often a tedious task and could potentially render a reproduction impossible.
While machine learning code is mostly distributed via GitHub, the power of Continuous Integration hasn't been used to its full potential. To solve this, Data Version Control (DVC) and Continuous Machine Learning (CML) have been introduced. DVC provides a language agnostic framework for managing parameters and results for the given machine learning pipelines together with GIT. The full potential of this setup can be found in the application of CML, which utilizes the complete CI/CD framework. Experiments are automatically submitted to provided computational resources through a single git commit. Furthermore, CML can be used to produce reports on these experiments. Finally, a complete overview is available through Iterative Studio. This allows for easier reproduction and collaboration on experiments.
