rng.uniform(low=0.0, high=1.0, size=static_input_shape).astype(
            np.float32
        )
    )

    def data_gen() -> repr_dataset.RepresentativeDataset:
      for _ in range(100):
        yield {
            'input_tensor': rng.uniform(
                low=0.0, high=1.0, size=static_input_shape
            ).astype(np.float32)
        }

    dataset_path = self.create_tempfile('tfrecord').full_path

module {
func.func @simpleWhile(%arg0: tensor<i32>) -> tensor<i32> {
  %0 = "tfl.while"(%arg0) ({
  ^bb0(%block: tensor<i32>):
    "tfl.yield"(%block) : (tensor<i32>) -> ()
  },{
  ^bb0(%block: tensor<i32>):
    "tfl.yield"(%block) : (tensor<i32>) -> ()
  }) {tac.device = "CPU", fused_activation_function = "RELU6", tac.inference_type = "FLOAT"} : (tensor<i32>) -> tensor<i32>
  func.return %0 : tensor<i32>
}

  regions accept the current value of the iteration variables as inputs.

  The condition region yields a tensor<i1> which, if false, will exit the loop.
  It can also, optionally and additionally, yield the iteration variables, which
  must be unchanged.

  The body region always has to yield the (possibly updated) iteration variables.

  The iteration variables are initialized to the Op input, and the results of the

		c := field.Name[0]
		if 'a' <= c && c <= 'z' || c == '_' {
			panic("reflect.StructOf: field \"" + field.Name + "\" is unexported but missing PkgPath")
		}
	}

	resolveReflectType(field.Type.common()) // install in runtime
	f := structField{
		Name:   newName(field.Name, string(field.Tag), field.IsExported(), field.Anonymous),
		Typ:    field.Type.common(),
		Offset: 0,
	}
	return f, field.PkgPath
}

          /*device_ordinal=*/builder.getI64IntegerAttr(0), device_type_attr),
      builder);
}

// Clones an IfRegionOp 'if_region' and attributes and creates then/else regions
// with yield op and an empty block.
mlir::TF::IfRegionOp CloneEmptyIfWithPredicate(mlir::TF::IfRegionOp if_region,
                                               OpBuilder& builder) {

// profile that event by storing a pointer to the lock in its "to be profiled
// upon unlock" field. If that field is already set, it uses the relative
// magnitudes to weight a random choice between itself and the other lock, with
// the loser's time being added to the "additional contention" field. Otherwise
// if the M's call stack buffer is occupied, it does the comparison against that
// sample's magnitude.
//

  region.push_back(new mlir::Block);
  auto saved_pos = op_builder.saveInsertionPoint();
  op_builder.setInsertionPointToEnd(&region.front());
  mlir::Operation* cloned_op = op_builder.clone(*op);
  // Add the yield operation.
  op_builder.create<mlir::TFL::YieldOp>(op_loc, cloned_op->getResults());
  // Now emit into the function body again.
  op_builder.restoreInsertionPoint(saved_pos);

		// is the respective default type.
		check.updateExprType(x.expr, Default(x.typ), true)
		check.updateExprType(y.expr, Default(y.typ), true)
	}

	// spec: "Comparison operators compare two operands and yield
	//        an untyped boolean value."
	x.typ = Typ[UntypedBool]
	return

Error:
	// We have an offending operand errOp and possibly an error cause.
	if cause == "" {

    %7 = "tf.FakeQuantWithMinMaxVarsPerChannel"(%arg1, %arg2, %arg3) {device = "", narrow_range = false, num_bits = 8 : i64} : (tensor<*xf32>, tensor<*xf32>, tensor<*xf32>) -> tensor<*xf32>
    "tfl.yield"(%7) : (tensor<*xf32>) -> ()
  }) {device = "", narrow_range = false, num_bits = 8 : i64} : (tensor<3x4xf32>, tensor<4xf32>, tensor<4xf32>) -> tensor<3x4xf32>

		// is the respective default type.
		check.updateExprType(x.expr, Default(x.typ), true)
		check.updateExprType(y.expr, Default(y.typ), true)
	}

	// spec: "Comparison operators compare two operands and yield
	//        an untyped boolean value."
	x.typ = Typ[UntypedBool]
	return

Error:
	// We have an offending operand errOp and possibly an error cause.
	if cause == "" {