hlo::getBroadcastDimensionsAttr(&rewriter, scaled, op.getStart()));
    return success();
  }
};

/// Converts a generic OpTy tensorflow op to a mhlo.reduce op over
/// ReductionOp.
/// `is_accumulation` controls whether it uses higher precision for the actual
/// reduction. This is set to false for ops like max where there is no precision
/// concerns.
//