Speaker
Raulian-Ionut Chiorescu
Description
This session will focus on available ML techniques for distributed training of models, hyperparameter optimization and model service. In particular, starting from a well known use case it will demonstrate:
- How to go from a script, to a docker image training on a single node, to a distributed training setup with multiple nodes
- How to do hyperparameter optimization, which kind of optimizers are available, how to monitor the workloads and how to publish the models
- How to serve models in production, at scale, with a simple http entrypoint or embedding the model in an application
