On September 21, 2023, the Broad Institute's Data Sciences Platform and Google's AI in Healthcare and Life Sciences team held a joint workshop, showcasing Terra capabilities using machine learning tools. The workshop included a hands-on demonstration of leveraging Terra for Vertex model training and deployment, and covered topics including GPU accelerators for managed training jobs, multi-node distributed training, and hyperparameter optimization.
You can find the slide deck containing links to a variety of tutorial and onboarding resources at broad.io/vertex-september2023 and you can check out the workspace used for the demonstration here: ML-PCAM-Classification-Vertex-AI
You can also find some helpful resources on Vertex AI and BigQuery in this supplemental article.
If you missed the workshop or would like to refresh your memory, please see the recording below, or you can find it on our YouTube channel as well.
Part 1: Train in a Terra Notebook
Learn how to get started training an ML model directly in Terra from a notebook
running in a Cloud Environment.
Part 2: Scaling, Training and Serving using Vertex AI
Learn how to run a training job using Google Cloud fully-managed services. You
don’t need to worry about the notebook environment automatically timing out during
a long training run.
Part 3: Hyper Parameter Optimization and Multi-node training
Learn how to how to run a hyperparameter tuning job on Vertex AI, using the
training code. We define the parameters that we want to vary during the HP search