tff.learning.LoopImplementation

An enum to specify which implementation of the training loop to use.

Attributes
`DATASET_ITERATOR`	A training loop that uses Dataset iterator ops. This is required when running on hosts with multiple GPUs to effectively leverage all GPUs on the host.
`DATASET_REDUCE`	A training loop that uses the DatasetReduce op. This is required when running on hosts with TPUs to allow the MLIR Bridge to compile the reduction function to XLA HLO for TPUs.

Class Variables
DATASET_ITERATOR	`<LoopImplementation.DATASET_ITERATOR: 1>`
DATASET_REDUCE	`<LoopImplementation.DATASET_REDUCE: 2>`

Except as otherwise noted, the content of this page is licensed under the Creative Commons Attribution 4.0 License, and code samples are licensed under the Apache 2.0 License. For details, see the Google Developers Site Policies. Java is a registered trademark of Oracle and/or its affiliates.

Last updated 2024-09-20 UTC.

Attributes