*   Add Conv3DTranspose as a builtin op.
*   `tf.summary`:
    *   Fix `tf.summary.should_record_summaries()` so it correctly reflects when
        summaries will be written, even when `tf.summary.record_if()` is not n
        effect, by returning True tensor if default writer is present.
*   Grappler:
    *   Disable default Grappler optimization timeout to make the optimization

}

def TF_WriteTrainingPredictionsOp : TF_Op<"WriteTrainingPredictions", [DeclareOpInterfaceMethods<MemoryEffectsOpInterface>]> {
  let summary = [{
Writes the given predictions into a RecordIO file using a previously
  }];

  let description = [{
initialized global TrainingPredictionWriter. The predictions are transformed
into a PredictionData proto before they are written to the file.
  }];