android-12.0.0_r34/s

syntax = "proto3";

package tensorflow.data;

// Represents the type of auto-sharding we enable.
enum AutoShardPolicy {
  AUTO = 0;
  FILE = 1;
  DATA = 2;
  OFF = -1;
}

message DistributeOptions {
  // The type of sharding that auto-shard should attempt. If this is set to
  // FILE, then we will attempt to shard by files (each worker will get a set of
  // files to process). If we cannot find a set of files to shard for at least
  // one file per worker, we will error out. When this option is selected, make
  // sure that you have enough files so that each worker gets at least one file.
  // There will be a runtime error thrown if there are insufficient files. If
  // this is set to DATA, then we will shard by elements produced by the
  // dataset, and each worker will process the whole dataset and discard the
  // portion that is not for itself. If this is set to OFF, then we will not
  // autoshard, and each worker will receive a copy of the full dataset. This
  // option is set to AUTO by default, AUTO will attempt to first shard by FILE,
  // and fall back to sharding by DATA if we cannot find a set of files to
  // shard.
  AutoShardPolicy auto_shard_policy = 1;
  // The number of devices attached to this input pipeline.
  oneof optional_num_devices {
    int32 num_devices = 2;
  }
}

message MapVectorization {
  // Whether to vectorize map transformations.
  oneof optional_enabled {
    bool enabled = 1;
  }
  // Whether to use ChooseFastestBranchDataset with this transformation. If
  // True, the pipeline picks between the vectorized and original segment at
  // runtime based on their iterations speed.
  oneof optional_use_choose_fastest {
    bool use_choose_fastest = 2;
  }
}

message OptimizationOptions {
  // Whether to apply default graph optimizations. If False, only graph
  // optimizations that have been explicitly enabled will be applied.
  oneof optional_apply_default_optimizations {
    bool apply_default_optimizations = 1;
  }
  // Whether to automatically tune performance knobs.
  oneof optional_autotune {
    bool autotune = 2;
  }
  // When autotuning is enabled (through autotune), determines whether to also
  // autotune buffer sizes for datasets with parallelism.
  oneof optional_autotune_buffers {
    bool autotune_buffers = 3;
  }
  // When autotuning is enabled (through autotune), determines the CPU budget to
  // use. Values greater than the number of schedulable CPU cores are allowed
  // but may result in CPU contention.
  oneof optional_autotune_cpu_budget {
    int32 autotune_cpu_budget = 4;
  }
  // When autotuning is enabled (through autotune), determines the RAM budget to
  // use. Values greater than the available RAM in bytes may result in OOM. If
  // 0, defaults to half of the available RAM in bytes.
  oneof optional_autotune_ram_budget {
    int32 autotune_ram_budget = 5;
  }
  // Whether to fuse filter transformations.
  oneof optional_filter_fusion {
    bool filter_fusion = 6;
  }
  // Whether to fuse filter dataset that predicts random_uniform < rate into a
  // sampling dataset.
  oneof optional_filter_with_random_uniform_fusion {
    bool filter_with_random_uniform_fusion = 7;
  }
  // Whether to hoist tf.random_uniform() ops out of map transformations.
  oneof optional_hoist_random_uniform {
    bool hoist_random_uniform = 8;
  }
  // Whether to fuse map and batch transformations.
  oneof optional_map_and_batch_fusion {
    bool map_and_batch_fusion = 9;
  }
  // Whether to fuse map and filter transformations.
  oneof optional_map_and_filter_fusion {
    bool map_and_filter_fusion = 10;
  }
  // Whether to fuse map transformations.
  oneof optional_map_fusion {
    bool map_fusion = 11;
  }
  // Whether to parallelize stateless map transformations.
  oneof optional_map_parallelization {
    bool map_parallelization = 12;
  }
  // The map vectorization options associated with the dataset.
  MapVectorization map_vectorization = 13;
  // Whether to eliminate no-op transformations.
  oneof optional_noop_elimination {
    bool noop_elimination = 14;
  }
  // Whether to parallelize copying of batch elements. This optimization is
  // highly experimental and can cause performance degradation (e.g. when the
  // parallelization overhead exceeds the benefits of performing the data copies
  // in parallel). You should only enable this optimization if a) your input
  // pipeline is bottlenecked on batching and b) you have validated that this
  // optimization improves performance.
  oneof optional_parallel_batch {
    bool parallel_batch = 15;
  }
  // Whether to reorder ops that will discard data to the front of unary
  // cardinality preserving transformations, e.g. dataset.map(...).take(3) will
  // be optimized to dataset.take(3).map(...). For now this optimization will
  // move `skip`, `shard` and `take` to the front of `map` and `prefetch`. This
  // optimization is only for performance; it will not affect the output of the
  // dataset.
  oneof optional_reorder_data_discarding_ops {
    bool reorder_data_discarding_ops = 16;
  }
  // Whether to fuse shuffle and repeat transformations.
  oneof optional_shuffle_and_repeat_fusion {
    bool shuffle_and_repeat_fusion = 17;
  }
}

message ThreadingOptions {
  // If set, it overrides the maximum degree of intra-op parallelism.
  oneof optional_max_intra_op_parallelism {
    int32 max_intra_op_parallelism = 1;
  }
  // If set, the dataset will use a private threadpool of the given size.
  oneof optional_private_threadpool_size {
    int32 private_threadpool_size = 2;
  }
}

// Represents how to handle external state during serialization.
enum ExternalStatePolicy {
  WARN = 0;
  IGNORE = 1;
  FAIL = 2;
}

// Message stored with Dataset objects to control how datasets are processed and
// optimized.
message Options {
  // Whether the outputs need to be produced in deterministic order.
  oneof optional_deterministic {
    bool deterministic = 1;
  }
  // The distribution strategy options associated with the dataset.
  DistributeOptions distribute_options = 2;
  // The optimization options associated with the dataset.
  OptimizationOptions optimization_options = 3;
  // Whether to introduce 'slack' in the last `prefetch` of the input pipeline,
  // if it exists. This may reduce CPU contention with accelerator host-side
  // activity at the start of a step. The slack frequency is determined by the
  // number of devices attached to this input pipeline.
  oneof optional_slack {
    bool slack = 4;
  }
  // The threading options associated with the dataset.
  ThreadingOptions threading_options = 5;
  // This option can be used to override the default policy for how to handle
  // external state when serializing a dataset or checkpointing its iterator.
  // There are three settings available - IGNORE: External state is ignored
  // without a warning; WARN: External state is ignored and a warning is logged;
  // FAIL: External state results in an error.
  oneof optional_external_state_policy {
    ExternalStatePolicy external_state_policy = 6;
  }
}