User guides

• Running basic experiments
  • Run independent trials in parallel
  • How does Tune compare to Ray Core tasks?
• Logging and outputs in Tune
  • How to configure logging in Tune?
  • How to log your Tune runs to TensorBoard?
  • How to control console output with Tune?
  • How to redirect trainable logs to files in a Tune run?
  • How do you log arbitrary files from a Tune trainable?
  • How to build custom Tune loggers?
• Setting trial resources
  • Use GPUs in Tune
  • Run distributed training with Tune
  • Limit concurrency in Tune
• Using search spaces
  • Use custom and conditional search spaces
• How to define stopping criteria for a Ray Tune experiment
  • Stop a Tune experiment manually
  • Stop with metric-based criteria
  • Stop after a time limit
  • Stop on trial failures
  • Stop trials early with Tune schedulers
  • Summary
• How to save and load trial checkpoints
  • Save and load checkpoints with the function API
  • Save and load checkpoints with the class API
  • Configure checkpointing
  • Summary
  • Appendix: Types of data stored by Tune
• Configure persistent storage in Ray Tune
  • Storage options in Tune
  • Examples
  • Advanced configuration
• How to enable fault tolerance in Ray Tune
  • Experiment-level fault tolerance in Tune
  • Trial-level fault tolerance in Tune
  • Summary
• Using callbacks and metrics
  • Use callbacks in Tune
  • Log metrics in Tune
• Get data in and out of Tune
  • Get data into Tune
  • Get data out of Tune
• Analyzing Tune Experiment Results
  • Loading experiment results from an directory
  • Experiment-level Analysis: Working with ResultGrid
  • Trial-level Analysis: Working with an individual Result
  • Plotting metrics
  • Accessing checkpoints and loading for test inference
  • Summary
• A Guide to Population Based Training with Tune
  • Function API with Population Based Training
  • Replaying a PBT run
  • Example: DCGAN with PBT
  • Visualization
  • Summary
• Deploying Tune in the cloud
  • Summary
  • Example: Distributed Tune on AWS VMs
  • Run a distributed Tune experiment
  • Configure storage for a distributed Tune run
  • Run Tune on preemptible instances
  • How does Tune handle failures?
  • Common Tune commands
  • Troubleshooting
• Tune architecture
  • What happens in Tuner.fit?
  • What’s the lifecycle of a trial?
  • What’s Tune’s architecture?
• Scalability benchmarks
  • Result throughput
  • Network overhead in Ray Tune