// CHECK-LABEL: testWhileRegionExternInCond
func.func @testWhileRegionExternInCond(%arg0 : tensor<*xf32>, %arg1 : tensor<i32>, %arg2 : tensor<i32>) -> tensor<*xf32> {
  %cst = arith.constant dense<4> : tensor<i32>
  %limit = "tf.Add"(%arg2, %cst) : (tensor<i32>, tensor<i32>) -> tensor<i32>
  // CHECK: [[Result:%.*]]:3 = "tf.While"(%arg0, %arg1, %{{.+}} <{body = @tf.WhileRegion_body, cond = @tf.WhileRegion_cond

// CHECK:         }

// -----

func.func @ensureBiasForConv2d(%arg0: tensor<128x32x32x3xf32>, %arg1: tensor<32x1x1x3xf32>) -> tensor<128x32x32x32xf32> {
  %cst = "tfl.no_value"() {value = unit} : () -> none

    module attributes {tf.versions = {bad_consumers = [], min_consumer = 0 : i32, producer = 268 : i32}} {
      func.func @main() -> (tensor<2xi32>) {
        %cst = "tf.Const"() {value = dense<[524170, 523952]> : tensor<2xi32>} : () -> tensor<2xi32>
        return %cst : tensor<2xi32>
    }
  })";

  auto compilation_result = CompileMlirModule(
      kHasReturnValuesAndNoMetadataRetvals,

  // CHECK: tf.VarHandleOp
  %0 = "tf.VarHandleOp"() {container = "", shared_name = "x"} : () -> tensor<!tf_type.resource<tensor<i32>>>
  %cst = "tf.Const"() {device = "/CPU:0", value = dense<true> : tensor<i1>} : () -> tensor<i1>
  // CHECK: "tf.If"(%cst, %0)
  %1 = "tf.If"(%cst, %0) {then_branch = @some_func, else_branch = @some_func, is_stateless = false} : (tensor<i1>, tensor<!tf_type.resource<tensor<i32>>>) -> tensor<i32>

      // can't lower tf.Const.
      Value cst;
      if (use_mhlo_const) {
        cst = builder.create<mhlo::ConstantOp>(func_op->getLoc(), attrs);
      } else {
        cst = builder.create<TF::ConstOp>(func_op->getLoc(), attrs);
      }
      func_op.getArgument(i).replaceAllUsesWith(cst);
    }
    return module_op;
  }

    %9 = "tf.Const"() <{value = dense<"test"> : tensor<3x!tf_type.string>}> : () -> tensor<3x!tf_type.string>
    %cst = "tf.Const"() <{value = dense<0> : tensor<i32>}> : () -> tensor<i32>
    %11:4 = "tf.Split"(%cst, %0) {num_split = 4 : i32} : (tensor<i32>, tensor<128x1024xf32>) -> (tensor<32x1024xf32>, tensor<32x1024xf32>, tensor<32x1024xf32>, tensor<32x1024xf32>)

			// We found a function that was executed.
			nCtrs := sd[i+coverage.NumCtrsOffset].Load()
			pkgId := sd[i+coverage.PkgIdOffset].Load()
			funcId := sd[i+coverage.FuncIdOffset].Load()
			cst := i + coverage.FirstCtrOffset
			counters := sd[cst : cst+int(nCtrs)]

			// Check to make sure that we have at least one live
			// counter. See the implementation note in ClearCoverageCounters
			// for a description of why this is needed.

  getOperation().walk([&](mlir::TF::ConstOp cst) {
    auto dict = cst->getAttrDictionary();
    if (dict.size() == 1) {
      return;  // Optimization. Assume the one attribute is "value".
    }
    // Recreate the attributes dictionary to only contain "value".
    NamedAttrList attributes;
    attributes.append(NamedAttribute(value_str_attr, cst->getAttr("value")));
    cst->setAttrs(attributes.getDictionary(&getContext()));

            }

            @Override
            public void visitLdcInsn(Object cst) {
                super.visitLdcInsn(remap(cst));
            }

            @Override
            public void visitLocalVariable(String name, String desc, String signature, Label start, Label end, int index) {

  %cst = "tf.Const"() {value = dense<1.1> : tensor<100x512xf32>} : () -> tensor<100x512xf32>
  %one = "tf.Const"() {value = dense<1> : tensor<i32>} : () -> tensor<i32>