Standard Names to Use for Graph Collections

The standard library uses various well-known names to collect and retrieve values associated with a graph.



For example, the tf$Optimizer subclasses default to optimizing the variables collected undergraph_keys()$TRAINABLE_VARIABLES if NULL is specified, but it is also possible to pass an explicit list of variables.

The following standard keys are defined:

  • GLOBAL_VARIABLES: the default collection of Variable objects, shared across distributed environment (model variables are subset of these). See tf$global_variables for more details. Commonly, all TRAINABLE_VARIABLES variables will be in MODEL_VARIABLES, and all MODEL_VARIABLES variables will be in GLOBAL_VARIABLES.

  • LOCAL_VARIABLES: the subset of Variable objects that are local to each machine. Usually used for temporarily variables, like counters. Note: use tf$contrib$framework$local_variable to add to this collection.

  • MODEL_VARIABLES: the subset of Variable objects that are used in the model for inference (feed forward). Note: use tf$contrib$framework$model_variable to add to this collection.

  • TRAINABLE_VARIABLES: the subset of Variable objects that will be trained by an optimizer. See tf$trainable_variables for more details.

  • SUMMARIES: the summary Tensor objects that have been created in the graph. See tf$summary$merge_all for more details.

  • QUEUE_RUNNERS: the QueueRunner objects that are used to produce input for a computation. See tf$train$start_queue_runners for more details.

  • MOVING_AVERAGE_VARIABLES: the subset of Variable objects that will also keep moving averages. See tf$moving_average_variables for more details.

  • REGULARIZATION_LOSSES: regularization losses collected during graph construction. The following standard keys are defined, but their collections are not automatically populated as many of the others are:


    • BIASES


See also

Other utility functions: latest_checkpoint


# }