// to be unique in the store.
//
// The storage interface presented here assumes that the underlying storage
// layer supports _Get_ (list), _Update_ (update), _Create_ (create) and
// _Delete_ semantics but does not guarantee any transactional semantics.
//
// _Update_, _Create_, and _Delete_ are mutator operations. These operations
// are asynchronous, and you might not see the effect immediately (e.g. _Get_

Bootstrapping can either be used to create a file from the beginning, or also to _update_ an existing file with new information.
Therefore, it's recommended to always use the same parameters once you started bootstrapping.

The dependency verification file can be generated with the following CLI instructions:

Tim Hockin <******@****.***> 1705105403 -0800

Tim Hockin <******@****.***> 1704583440 -0800

        | xargs -0 rm -f

    if kube::protoc::check_protoc >/dev/null; then
      hack/_update-generated-protobuf-dockerized.sh "${apis[@]}"
    else
      kube::log::status "protoc ${PROTOC_VERSION} not found (can install with hack/install-protoc.sh); generating containerized..."
      build/run.sh hack/_update-generated-protobuf-dockerized.sh "${apis[@]}"
    fi
}

# Deep-copy generation
#

}
```

This way, the first convolution with 3 feature dimension will be transformed
to 12 feature dimension, which has better performance on TPU.
### `-tf-tpu-update-embedding-enqueue-op-inputs`

_Updates inputs to TPU embedding enqueue ops depending on whether graph is in training mode or in evaluation mode._

Updates inputs to TPU embedding enqueue ops depending on whether graph
is in training mode or in evaluation mode.