def : Pat<(TF_RandomUniformOp:$old $shape, $seed, $seed2),
          (MHLO_RngOp
            (MHLO_ConstantOp
              (NativeCodeCall<"$_builder.getFloatAttr(old.getDtype(), 0.0)">)),
            (MHLO_ConstantOp
              (NativeCodeCall<"$_builder.getFloatAttr(old.getDtype(), 1.0)">)),
            (CastValueToI64 $old, $shape),
            MHLO_RngDistributionValue<"UNIFORM">),
          [(IsShapedTensor $shape)]>;

  sources.reserve(output_tensors.size());
  for (const TapeTensor& output_tensor : output_tensors) {
    // Ownership of `aid` transferred to CallBackwardFunction below.
    Gradient* aid;
    if (output_tensor.GetDType() == tensorflow::DT_VARIANT) {
      // Note: Needs to be zeros rather than ones since there's currently no
      // ones_like for variants.
      aid = output_tensor.ZerosLike();
    } else {

  handle_ = other.handle_;
  handle_->Ref();
}
TapeTensor::~TapeTensor() { handle_->Unref(); }

int64_t TapeTensor::GetID() const { return ToId(handle_); }

tensorflow::DataType TapeTensor::GetDType() const {
  return handle_->DataType();
}
AbstractTensorHandle* TapeTensor::GetHandle() const { return handle_; }

AbstractTensorHandle* TapeTensor::ZerosLike() const { return nullptr; }

class TapeVSpace

                CallableInfo callable = request.getInterceptedCallable();
                CallableKindInfo kind = callable.getKind();
                if (kind == CallableKindInfo.AFTER_CONSTRUCTOR) {
                    Type constructedType = request.getInterceptedCallable().getOwner().getType();
                    String typeKey = implementationName + ":" + constructedType;

    assertTrue(
        from.getType() + " is expected to be assignable to " + to.getType(),
        to.isSupertypeOf(from));
    assertTrue(
        to.getType() + " is expected to be a supertype of " + from.getType(),
        to.isSupertypeOf(from));
    assertTrue(
        from.getType() + " is expected to be a subtype of " + to.getType(), from.isSubtypeOf(to));
  }

    private static String getFqName(CallInterceptionRequest request) {
        String propertyName = request.getRequestExtras().getByType(PropertyUpgradeRequestExtra.class).get().getPropertyName();
        String containingType = request.getInterceptedCallable().getOwner().getType().getClassName();
        return containingType + "#" + propertyName;
    }

  auto result_type =
      OpTrait::util::getBroadcastedType(x.getType(), y.getType());
  if (!result_type) {
    if (incompatible_shape_error.getValue()) {
      mlir::emitError(loc, "non-broadcastable operands");
    } else {
      return UnrankedTensorType::get(builder->getI1Type());
    }
  }

  auto ranked_type = mlir::dyn_cast<RankedTensorType>(result_type);

  LogicalResult matchAndRewrite(TFL::GatherOp op,
                                PatternRewriter &rewriter) const override {
    auto const_type = mlir::dyn_cast<ShapedType>(op.getOperand(0).getType());
    auto result_type = mlir::dyn_cast<ShapedType>(op.getResult().getType());
    if (!const_type || !const_type.getElementType().isSignlessInteger(4) ||
        !result_type || !result_type.getElementType().isSignlessInteger(8)) {
      return failure();
    }

                  int32_t dim) {
  Type attribute_type = builder.getI64Type();
  return builder.create<TF::StridedSliceOp>(
      loc,
      RankedTensorType::get(
          {}, mlir::cast<ShapedType>(shape_value.getType()).getElementType()),
      /*input=*/shape_value,
      /*begin=*/Create1DConstValue<int32_t>(builder, loc, {dim}),
      /*end=*/Create1DConstValue<int32_t>(builder, loc, {dim + 1}),

                .addModifiers(Modifier.PUBLIC, Modifier.STATIC)
                .addParameter(typeName(callable.getOwner().getType()), SELF_PARAMETER_NAME)
                .addParameters(parameters)
                .addCode(generateMethodBody(implementation, callable, implementationExtra))
                .returns(typeName(callable.getReturnType().getType()))
                .addAnnotations(getAnnotations(implementationExtra))
                .build();