if (!mlir::isa<FloatType>(layerStatsType.getElementType())) {
      return emitOpError("layerStats must have a floating point element type");
    }
    if (layerStatsType.getRank() != 1 || layerStatsType.getDimSize(0) != 2) {
      return emitOpError("layerStats must have shape [2]");
    }
  }
  // Verify axisStats (optional) attribute.
  if (getAxisStats()) {

	case float64:
		expiration = time.Unix(int64(exp), 0)
	case json.Number:
		v, _ := exp.Int64()
		expiration = time.Unix(v, 0)
	}
	return expiration, nil
}

// GetAud returns the claim `aud` from the token. Returns nil if not found.
func GetAud(token string) ([]string, error) {
	claims, err := parseJwtClaims(token)
	if err != nil {
		return nil, err
	}

	rawAud := claims["aud"]
	if rawAud == nil {

  // If the original type doesn't have a rank, then refine as the updated type
  // has a rank.
  if (!original.hasRank()) return true;

  // Both types must have the same rank.
  if (original.getRank() != updated.getRank()) return false;

  // Refine if the updated type is bounded.
  return IsBounded(updated);
}

// Propagates more refined type by cloning op using the new operands. This

  }

  // Currently we only support compressing weights of ops:
  //   Conv, DepthwiseConv, TransposeConv, whose filter has rank 4, and
  //   FullyConnected, whose filter has rank 2.
  if (type.getRank() != 2 && type.getRank() != 4) {
    result.can_compress = false;
    return result;
  }

  float random_sparsity = CalculateRandomSparsity(attr, type);
  if (random_sparsity < kMinSparsityLevel) {

		},
		"two audiences list": {
			jwt: twoAudList,
			aud: []string{"abc", "xyz"},
		},
	}

	for id, tc := range testCases {
		t.Run(id, func(t *testing.T) {
			if got, _ := GetAud(tc.jwt); !reflect.DeepEqual(tc.aud, got) {
				t.Errorf("want audience %v but got %v", tc.aud, got)
			}
		})
	}
}

func TestBase64UrlPartDecoding(t *testing.T) {

			targetRef.GetKind() == gvk.KubernetesGateway.Kind &&
			target.Name == gatewayName &&
			(targetRef.GetNamespace() == "" || targetRef.GetNamespace() == p.Namespace) {
			return true
		}

		// Service attached
		if p.IsWaypoint &&
			config.CanonicalGroup(targetRef.GetGroup()) == gvk.Service.CanonicalGroup() &&
			targetRef.GetKind() == gvk.Service.Kind &&

inline Type TransposeLastTwoDims(Type type) {
  auto shaped_type = type.dyn_cast<ShapedType>();
  if (!shaped_type.hasStaticShape() || shaped_type.getRank() < 2) {
    return nullptr;
  }
  int rank = shaped_type.getRank();
  if (rank < 2) {
    return nullptr;
  }
  SmallVector<int64_t> new_shape(shaped_type.getShape().begin(),
                                 shaped_type.getShape().end());

  return lhs.hasRank() && rhs.hasRank() && lhs.getRank() == rhs.getRank();
}

// Creates a compatible RankedTensorType where mismatched dimensions are
// replaced with dynamic sizes.
RankedTensorType GetCompatibleRankedTensorType(RankedTensorType lhs,
                                               RankedTensorType rhs) {
  assert(lhs.getRank() == rhs.getRank());
  llvm::SmallVector<int64_t, 4> dims;

  // Output type is unranked if input type is not ranked.
  auto ranked_ty = mlir::dyn_cast<RankedTensorType>(input_ty);
  if (!ranked_ty) return UnrankedTensorType::get(element_ty);
  int64_t rank = ranked_ty.getRank();

  DenseIntElementsAttr indices;
  if (!matchPattern(reduction_indices, m_Constant(&indices))) {
    // Output type is unranked if reduction indices are not constant and reduced
    // dimensions are not kept.

    // Compute mean
    int64_t input_last_dim = input_type.getRank() - 1;
    auto dims_type = RankedTensorType::get(/*shape=*/{input_last_dim},
                                           rewriter.getIntegerType(32));
    ::mlir::SmallVector<int32_t> reduce_dim_axes;
    for (int i = 0; i < input_type.getRank(); ++i) {
      if (i != feature_index) {
        reduce_dim_axes.push_back(i);
      }
    }