Apache Beam and Ensemble Modeling: A Winning Combination for Machine Learning


Speaker(s):
Shubham Krishna

Apache Beam and Ensemble Modeling: A Winning Combination for Machine Learning

By Shubham Krishna
Jul-18 16:30-16:55 in A
Add to Calendar 07/18/2023 4:30 PM 07/18/2023 4:55 PM America/Los_Angeles AS24: Apache Beam and Ensemble Modeling: A Winning Combination for Machine Learning

Are you looking for ways to streamline your machine learning pipeline and make it more efficient? Look no further than this talk on Apache Beam and ensemble modeling. In this session, we’ll show how to leverage the power of Apache Beam’s flexible data processing framework and the RunInference API to simplify your workflow for complex machine learning tasks.

One of the biggest challenges in developing machine learning systems is managing the various steps involved in the process. From data ingestion to processing tasks, inference, and post-processing, there are a lot of moving parts to keep track of. But with Apache Beam’s single DAG (directed acyclic graph) encapsulation, you can orchestrate all of those steps together in a streamlined, efficient manner. This allows you to build resilient and scalable end-to-end machine learning systems.

We’ll demonstrate how you can use the RunInference API to deploy your machine learning model in a Beam pipeline. By integrating your model as a step in your DAG, you can compose multiple RunInference transforms within a single pipeline. This makes it easier than ever to build complex ML systems with multiple models.

We’ll walk you through an end-to-end example of an ensemble model pipeline used for generating and ranking image captions. Using two open-source models - the BLIP model for image caption generation and the CLIP model for caption ranking - we’ll show you how to implement a powerful image captioning system that ranks captions based on how well they describe the input image.

After attending this talk, you’ll have a deeper understanding of how Apache Beam and ensemble modeling can simplify your machine learning workflow and help you build more effective systems.

Session 25m: Live session of 25 minutes

 A

  Download slides

Are you looking for ways to streamline your machine learning pipeline and make it more efficient? Look no further than this talk on Apache Beam and ensemble modeling. In this session, we’ll show how to leverage the power of Apache Beam’s flexible data processing framework and the RunInference API to simplify your workflow for complex machine learning tasks.

One of the biggest challenges in developing machine learning systems is managing the various steps involved in the process. From data ingestion to processing tasks, inference, and post-processing, there are a lot of moving parts to keep track of. But with Apache Beam’s single DAG (directed acyclic graph) encapsulation, you can orchestrate all of those steps together in a streamlined, efficient manner. This allows you to build resilient and scalable end-to-end machine learning systems.

We’ll demonstrate how you can use the RunInference API to deploy your machine learning model in a Beam pipeline. By integrating your model as a step in your DAG, you can compose multiple RunInference transforms within a single pipeline. This makes it easier than ever to build complex ML systems with multiple models.

We’ll walk you through an end-to-end example of an ensemble model pipeline used for generating and ranking image captions. Using two open-source models - the BLIP model for image caption generation and the CLIP model for caption ranking - we’ll show you how to implement a powerful image captioning system that ranks captions based on how well they describe the input image.

After attending this talk, you’ll have a deeper understanding of how Apache Beam and ensemble modeling can simplify your machine learning workflow and help you build more effective systems.

Session 25m: Live session of 25 minutes