Faster model training and experimentation with Vertex AI
summary
Shows how to convert a TensorFlow training script into a containerized Vertex AI job that runs MirroredStrategy on two V100 GPUs, reports metrics with the Hypertune library, and launches a serverless hyperparameter tuning sweep via the Vertex AI Python SDK. Example builds a horses-vs-humans classifier, runs six trials (configurable) exploring learning rate, momentum and hidden units, returns a best validation accuracy ~0.82, and demonstrates that multi-GPU training plus automated tuning shortens iteration time and produces an actionable hyperparameter setting.
AI-written summary · about
chapters (11)
- 0:00Introduction to Vertex AI
- 2:39Vertex AI and use case
- 5:20Distributed training concepts
- 6:53Environment setup
- 9:57Preparing training code
- 13:24Implementing distributed training
- 15:22Hyperparameter tuning setup
- 19:16Containerizing the code
- 22:47Launching the tuning job
- 26:43Tuning strategy
- 29:17Reviewing tuning results
related talks
18mBuilding Efficient PyTorch Models on Vertex AI
51mHands on with Cloud AI
12mBuild end-to-end solutions with Vertex AI
45mVertex AI: Pipelines for your MLOps workflows
02mFuel your custom models on the Cloud AI Platform
26mBuilding Portable ML Pipelines for Development and Production
40mVertex AI: Pipelines for your MLOps workflows
31mGoogle's AI Map