#   will need to run `docker exec tf pkill bazel` to quit bazel.
#
#   Note: new files created from the container are owned by "root".
#   You can run e.g. `docker exec tf chown -R $(id -u):$(id -g) build_output`
#   to transfer ownership to your user.
#
# Docker is enabled by default on Linux. You may disable it if you prefer:
# export TFCI=py311,linux_x86,no_docker

# Advanced: Use Remote Build Execution (RBE) (internal developers only)

  // Configure graph collection in RunMetadata.
  virtual void SetShouldStoreGraphs(bool value) = 0;

  // Return the collected RunMetadata. This method will transfer the ownership
  // to the caller.
  virtual std::unique_ptr<RunMetadata> ExportRunMetadata() = 0;

  // For LLVM style RTTI.
  static bool classof(const AbstractContext* ptr) {

///   * `TF_InitPlugin` function: must be present in the plugin shared object as
///     it will be called by core TensorFlow when the filesystem plugin is
///     loaded;
///   * `TF_FilesystemPluginOps` struct: used to transfer information between
///     plugins and core TensorFlow about the operations provided and metadata;
///   * `TF_FilesystemPluginInfo` struct: similar to the above structure, but

build:rbe_base --remote_executor=grpcs://remotebuildexecution.googleapis.com
build:rbe_base --remote_timeout=3600
build:rbe_base --spawn_strategy=remote,worker,standalone,local
# Attempt to minimize the amount of data transfer between bazel and the remote
# workers:
build:rbe_base --remote_download_toplevel
test:rbe_base --test_env=USER=anon

# TODO(kanglan): Check if we want to merge rbe_linux into rbe_linux_cpu.

    distributing datasets. For this to work, the dispatcher's `work_dir` must be
    accessible from workers. If the worker fails to read from the `work_dir`, it
    falls back to using RPC for dataset graph transfer.
*   Adds support for a new "distributed_epoch" processing mode. This processing
    mode distributes a dataset across all tf.data workers, instead of having
    each worker process the full dataset. See