PopART User Guide

• 1. Introduction
• 2. Importing graphs
  • 2.1. Creating a session
  • 2.2. Session control options
• 3. Building graphs in PopART
  • 3.1. Adding operations to the graph
  • 3.2. Adding parameters to the graph
  • 3.3. Setting outputs
  • 3.4. Setting the IPU number for operations
• 4. Executing graphs
  • 4.1. Setting input/output data buffers for an execution
  • 4.2. Selecting a device for execution
  • 4.3. Executing a session
  • 4.4. Saving and loading a model
  • 4.5. Retrieving profiling reports
  • 4.6. Turning on execution tracing
• 5. Examples
• 6. Distributed training with Horovod
  • 6.1. How to modify a PopART program for distributed training
  • 6.2. Install
  • 6.3. Configuring and running distributed training
  • 6.4. Full distributed training example
• 7. Performance optimisation
  • 7.1. Sync configuration
• 8. Supported operators
  • 8.1. Domain: ai.graphcore
  • 8.2. Domain: ai.onnx
• 9. Environment variables
  • 9.1. Logging
  • 9.2. Generating DOT files
  • 9.3. Saving the tensor map
  • 9.4. Inspecting the Ir
• 10. References
• 11. Glossary
  • 11.1. Sample
  • 11.2. Micro-batch size
  • 11.3. Replication factor
  • 11.4. Accumulation factor
  • 11.5. Batch size
  • 11.6. Batches per step
  • 11.7. Step size
  • 11.8. Input data shape
• 12. Trademarks & copyright