Anyscale

Introduction

Imports & Loading the Dataset

Defining the Model

Train Loop Per Worker

Defining the Training Loop Configuration

Configuring the Scaling Config

Defining the Model Wrapper

Building the Dataloader

Reporting Metrics and Checkpointing

Persistent Storage

Putting it all together with TorchTrainer

Inspecting the Training Results

Inference with Your Trained Model

Full Chapter Notebook

Introduction

Train Loop Using Ray Data

Building a Ray Data-Backed Dataloader

Preparing and Loading the Dataset for Ray Data

Transformations with Ray Data

Configuring TorchTrainer and Launching Training

Full Chapter Notebook

Introduction

Checkpoint Loading for Fault Tolerance

Saving Fault Tolerant Checkpoints

Launching Fault Tolerant Training

Manually Restoring from Checkpoints

Cleaning up Cluster Storage and Conclusion

Concluding the Intro Tutorials and Next Steps

Full Chapter Notebook