Targeting the IPU from TensorFlow 2

• 1. Introduction
  • 1.1. Document overview
• 2. Targeting the Poplar XLA device
  • 2.1. Supported types
  • 2.2. Device selection
  • 2.3. Configuring system options
    • 2.3.1. TF_POPLAR_FLAGS environment variable
  • 2.4. Supported operations
  • 2.5. Unsupported operations
  • 2.6. Error Handling
    • 2.6.1. Construction and compilation errors
    • 2.6.2. Runtime errors
• 3. Support for TensorFlow 2
  • 3.1. IPUStrategy
  • 3.2. Execution modes
    • 3.2.1. Graph mode with @tf.function
    • 3.2.2. Eager mode
  • 3.3. On-device loops
• 4. Keras with IPUs
  • 4.1. Single IPU models
  • 4.2. Using steps_per_execution
  • 4.3. Gradient accumulation
  • 4.4. Model parallelism
    • 4.4.1. Sequential model
    • 4.4.2. Functional model
      • Pipelining a model you are writing yourself
      • Pipelining an existing functional model
  • 4.5. Automatic data parallelism
  • 4.6. Asynchronous callbacks
  • 4.7. Porting models from TensorFlow 2.1
    • 4.7.1. TF2.1
    • 4.7.2. TF2.4
  • 4.8. Implementation details
• 5. Compiling and pre-compiling executables
  • 5.1. Caching of compiled executables
  • 5.2. Pre-compiling executables
    • 5.2.1. Unsupported Operations
• 6. Training a model
  • 6.1. Training loops, data sets and feed queues
  • 6.2. Accessing outfeed queue results during execution
  • 6.3. Replicated graphs
    • 6.3.1. Selecting the number of replicas
    • 6.3.2. Performing parameter updates
  • 6.4. Pipelined training
    • 6.4.1. Sequential scheduling
    • 6.4.2. Interleaved scheduling
    • 6.4.3. Grouped scheduling
    • 6.4.4. Pipeline stage inputs and outputs
    • 6.4.5. Applying an optimiser to the graph
    • 6.4.6. Device mapping
    • 6.4.7. Concurrent pipeline stages
  • 6.5. Gradient accumulation
    • 6.5.1. Optimizers
    • 6.5.2. Pipelining
    • 6.5.3. Accumulation data type
  • 6.6. Optimizer state offloading
  • 6.7. Dataset benchmarking
    • 6.7.1. Accessing the JSON data
• 7. Efficient IPU I/O
  • 7.1. Prefetch elements
  • 7.2. I/O Tiles
• 8. Example using IPUEstimator
• 9. Example using IPUPipelineEstimator
• 10. Distributed training
  • 10.1. Example using IPUMultiWorkerStrategy
    • 10.1.1. The input function
    • 10.1.2. The model function
    • 10.1.3. Cluster definition
    • 10.1.4. Complete example
  • 10.2. Distributed training with Horovod
  • 10.3. Launching Horovod training
  • 10.4. Complete Horovod example
• 11. Half-precision floating point and stochastic rounding
  • 11.1. Controlling the half-precision floating-point unit
  • 11.2. Resetting the global random number seed
  • 11.3. Debugging numerical issues
• 12. IPU-optimised operations
  • 12.1. LSTM and GRU
  • 12.2. Dropout
  • 12.3. Embedding lookup
  • 12.4. Group normalisation
  • 12.5. Instance normalisation
  • 12.6. Layer normalisation
  • 12.7. GeLU activation
  • 12.8. Sequence slice
  • 12.9. Histogram
• 13. IPU Outlined Functions
  • 13.1. Usage
  • 13.2. Examples
    • 13.2.1. Models with common structures
    • 13.2.2. Serializing large operations
• 14. Writing custom operations
  • 14.1. Custom operation on the IPU
    • 14.1.1. Building the Poplar graph
    • 14.1.2. Gradient builders
    • 14.1.3. Metadata
    • 14.1.4. Compiling the IPU code
      • API level
      • PopLibs library code
      • Compiling the library file
    • 14.1.5. Using the custom op in TensorFlow
    • 14.1.6. Tensor allocation
    • 14.1.7. Examples
      • In-place operations
      • Operation attributes
      • Custom codelet
  • 14.2. Custom host CPU operations
    • 14.2.1. Gradient callback
• 15. IPU host embeddings
  • 15.1. Usage
  • 15.2. Example
  • 15.3. Experimental functionality: IPU embeddings in remote buffers
    • 15.3.1. Partitioning strategies
      • Token strategy
      • Encoding strategy
      • Choosing a strategy for your application
• 16. Retrieving information about compilation and execution
  • 16.1. TensorFlow options for reporting
  • 16.2. Dumping auxiliary Poplar information
    • 16.2.1. Poplar vertex graph
    • 16.2.2. Poplar interval report
  • 16.3. XLA graph file naming
• 17. API changes
  • 17.1. Release 2.2
    • 17.1.1. Breaking changes
      • C++ Poplar TensorFlow libraries are private by default
      • Reports removed from ipu events
      • TensorFlow 2.1 to TensorFlow 2.4 Migration
    • 17.1.2. Non-breaking changes
      • IPULoggingTensorHook replication_factor deprecated
      • IPUInfeedQueue/IPUOutfeedQueue/IPULoggingTensorHook feed_name deprecated
      • Change of output location for profiling information
  • 17.2. Release 2.1
    • 17.2.1. Breaking changes
      • IPUPipelineEstimator change
      • Autosharding removed
      • IPU config change
      • IPU Keras changes [TensorFlow 2]
    • 17.2.2. Non-breaking changes
      • Recompute suggestions deprecated
      • IPUInfeedQueue/IPUOutfeedQueue replication_factor deprecated
      • IPUInfeedQueue data_to_prefetch deprecated
      • IPUOutfeedQueue data_to_prefetch deprecated
      • CTC loss ops deprecated
      • New configuration API
      • Support for grouped collectives
      • Environment variable changes
  • 17.3. Release 2.0
    • 17.3.1. Breaking changes
    • 17.3.2. Non-breaking changes
      • IPUPipelineEstimator change
      • Autosharding deprecated
      • IPU config change
      • IPU Keras changes [TensorFlow 2]
• 18. Python API
  • 18.1. Operations and utilities related to the Graphcore IPU
  • 18.2. Distribution strategy for a single system
  • 18.3. Compiler interface
  • 18.4. Scoping contexts
  • 18.5. Infeed queue
  • 18.6. Outfeed queue
  • 18.7. General utilities
  • 18.8. Configuration utilities
  • 18.9. Looping utilities
  • 18.10. Distributed training
  • 18.11. Horovod
  • 18.12. Datasets
    • 18.12.1. Dataset benchmarking
    • 18.12.2. Dataset wrappers
  • 18.13. Estimators
    • 18.13.1. IPUEstimator
    • 18.13.2. IPUPipelineEstimator
    • 18.13.3. Run configs
    • 18.13.4. Session run hooks
  • 18.14. Keras
    • 18.14.1. IPU specific Keras extensions
  • 18.15. Keras layers
    • 18.15.1. Keras layer specializations for the Graphcore IPU
  • 18.16. Keras losses
    • 18.16.1. Keras loss functions for the Graphcore IPU
  • 18.17. Keras optimizers
    • 18.17.1. Keras Optimizer wrappers for the Graphcore IPU
  • 18.18. Operators
    • 18.18.1. Custom operations
    • 18.18.2. Functional operators
    • 18.18.3. Image operations
    • 18.18.4. Graphcore utility operations
    • 18.18.5. IPU specific maths operations
    • 18.18.6. Pipelining operators
    • 18.18.7. Popnn primitive neural network operators
    • 18.18.8. Popnn normalization operators
    • 18.18.9. Popnn recurrent neural network operators
    • 18.18.10. Popops all to all and all gather operators
    • 18.18.11. Popops cross replica operators
    • 18.18.12. Popops embedding operators
    • 18.18.13. Popops reduce scatter operator
    • 18.18.14. Poprand operators
    • 18.18.15. Utility operations to be used in replicated mode
    • 18.18.16. Slicing operators
    • 18.18.17. Statistics operators
    • 18.18.18. Summary operations for IPUs
  • 18.19. Optimisers
    • 18.19.1. Optimizer classes for the Graphcore IPU
  • 18.20. Sharding
    • 18.20.1. Utility functions for sharding graphs
• 19. TensorFlow operators supported by the IPU
• 20. Resources
  • 20.1. Graphcore
  • 20.2. TensorFlow
  • 20.3. Other
• 21. Index
• 22. Trademarks & copyright