ngraph/test/runtime/pass/opset0_downgrade.cpp

   1 //*****************************************************************************
   2 // Copyright 2017-2020 Intel Corporation
   3 //
   4 // Licensed under the Apache License, Version 2.0 (the "License");
   5 // you may not use this file except in compliance with the License.
   6 // You may obtain a copy of the License at
   7 //
   8 //     http://www.apache.org/licenses/LICENSE-2.0
   9 //
  10 // Unless required by applicable law or agreed to in writing, software
  11 // distributed under the License is distributed on an "AS IS" BASIS,
  12 // WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied.
  13 // See the License for the specific language governing permissions and
  14 // limitations under the License.
  15 //*****************************************************************************
  16
  17 #include <algorithm>
  18 #include <cstdint>
  19 #include <functional>
  20 #include <numeric>
  21
  22 #include "ngraph/builder/autobroadcast.hpp"
  23 #include "ngraph/builder/reshape.hpp"
  24 #include "ngraph/graph_util.hpp"
  25 #include "ngraph/node.hpp"
  26 #include "ngraph/op/util/attr_types.hpp"
  27 #include "ngraph/op/util/op_types.hpp"
  28 #include "ngraph/ops.hpp"
  29 #include "ngraph/provenance.hpp"
  30 #include "ngraph/slice_plan.hpp"
  31 #include "ngraph/type.hpp"
  32 #include "ngraph/validation_util.hpp"
  33 #include "op/avg_pool.hpp"
  34 #include "op/convolution.hpp"
  35 #include "op/group_conv.hpp"
  36 #include "pass/implicit_broadcast_elimination.hpp"
  37 #include "pass/opset0_downgrade.hpp"
  38
  39 NGRAPH_SUPPRESS_DEPRECATED_START
  40
  41 using namespace std;
  42 using namespace ngraph;
  43
  44 namespace opset0_downgrade
  45 {
  46     template <typename OpV0, typename OpV1>
  47     shared_ptr<Node> op_cast_binary_elementwise_node(const shared_ptr<OpV1>& node)
  48     {
  49         const auto input_arg0 = node->input_value(0);
  50         const auto input_arg1 = node->input_value(1);
  51         const auto autob = node->get_autob();
  52         auto replacement_node = make_shared<OpV0>(input_arg0, input_arg1, autob);
  53         replace_node(node, replacement_node);
  54         return replacement_node;
  55     }
  56
  57     template <typename OpV0, typename OpV1>
  58     shared_ptr<Node> op_cast_reduction_node(const shared_ptr<OpV1>& node)
  59     {
  60         auto replacement_node = make_shared<OpV0>(node->input_value(0), node->input_value(1));
  61         if (node->get_keep_dims())
  62         {
  63             string v1_op_name = string{node->get_type_name()} + ":v1";
  64             string v0_op_name = string{OpV0{}.get_type_name()} + ":v0";
  65
  66             NGRAPH_CHECK(node->reduction_axes_constant(),
  67                          "Unable to convert ",
  68                          v1_op_name,
  69                          "to ",
  70                          v0_op_name,
  71                          " if reduction axes are not constant (for keep_dims=true). Node: ",
  72                          *node);
  73             auto output_pshape = replacement_node->get_output_partial_shape(0);
  74             NGRAPH_CHECK(output_pshape.is_static(),
  75                          "Unable to convert ",
  76                          v1_op_name,
  77                          "to ",
  78                          v0_op_name,
  79                          " if output shape is dynamic (for keep_dims=true). Node: ",
  80                          *node);
  81             const auto output_shape = output_pshape.to_shape();
  82             auto reshaped_output_shape = output_shape;
  83             for (const auto& axis : node->get_reduction_axes())
  84             {
  85                 reshaped_output_shape.insert(reshaped_output_shape.begin() + axis, 1);
  86             }
  87             auto reshaped_product = make_shared<op::Reshape>(replacement_node->output(0),
  88                                                              get_default_order(output_shape),
  89                                                              reshaped_output_shape);
  90             return reshaped_product;
  91         }
  92         else
  93         {
  94             return replacement_node;
  95         }
  96     }
  97
  98     // Default is that we did nothing
  99     shared_ptr<Node> op_cast(shared_ptr<Node> node) { return nullptr; }
 100     shared_ptr<Node> op_cast(shared_ptr<op::v1::Add> node)
 101     {
 102         return op_cast_binary_elementwise_node<op::v0::Add, op::v1::Add>(node);
 103     }
 104
 105     shared_ptr<Node> op_cast(shared_ptr<op::v1::AvgPool> node)
 106     {
 107         auto const input_arg = node->input_value(0);
 108         const auto ceil_mode = static_cast<bool>(node->get_rounding_type());
 109         const auto include_padding_in_avg_computation = !node->get_exclude_pad();
 110         const auto pad_type = node->get_auto_pad();
 111         const auto padding_below = node->get_pads_begin();
 112         const auto padding_above = node->get_pads_end();
 113         const auto window_movement_strides = node->get_strides();
 114         const auto window_shape = node->get_kernel();
 115
 116         auto replacement_node = make_shared<op::v0::AvgPool>(input_arg,
 117                                                              window_shape,
 118                                                              window_movement_strides,
 119                                                              padding_below,
 120                                                              padding_above,
 121                                                              include_padding_in_avg_computation,
 122                                                              pad_type,
 123                                                              ceil_mode);
 124         replace_node(node, replacement_node);
 125         return replacement_node;
 126     }
 127
 128     shared_ptr<Node> op_cast(shared_ptr<op::v1::Convolution> node)
 129     {
 130         const auto data_arg = node->input_value(0);
 131         const auto filters_arg = node->input_value(1);
 132         const auto strides = node->get_strides();
 133         const size_t num_spatial_dims = strides.size();
 134         auto replacement_node = make_shared<op::v0::Convolution>(data_arg,
 135                                                                  filters_arg,
 136                                                                  node->get_strides(),
 137                                                                  node->get_dilations(),
 138                                                                  node->get_pads_begin(),
 139                                                                  node->get_pads_end(),
 140                                                                  Strides(num_spatial_dims, 1),
 141                                                                  node->get_auto_pad());
 142         replace_node(node, replacement_node);
 143         return replacement_node;
 144     }
 145
 146     shared_ptr<Node> op_cast(shared_ptr<op::v1::ConvolutionBackpropData> node)
 147     {
 148         const auto data_arg = node->input_value(0);
 149         const auto filters_arg = node->input_value(1);
 150
 151         auto data_pshape = data_arg.get_partial_shape();
 152         auto filters_pshape = filters_arg.get_partial_shape();
 153
 154         NGRAPH_CHECK(data_pshape.rank().is_static() && data_pshape[0].is_static() &&
 155                          filters_pshape.rank().is_static() && filters_pshape[1].is_static(),
 156                      "Unable to convert ConvolutionBackpropData:v1 to ConvolutionBackpropData:v0 "
 157                      "if data shape N and filters shape C dimensions are not static. Node: ",
 158                      *node);
 159
 160         const size_t num_spatial_dims = data_pshape.rank().get_length() - 2;
 161
 162         const PartialShape output_pshape{node->get_output_partial_shape(0)};
 163         NGRAPH_CHECK(output_pshape.is_static(),
 164                      "Unable to convert ConvolutionBackpropData:v1 to ConvolutionBackpropData:v0 "
 165                      "if output shape is dynamic. Node: ",
 166                      *node);
 167         Shape output_shape = output_pshape.to_shape();
 168
 169         auto replacement_node =
 170             make_shared<op::v0::ConvolutionBackpropData>(output_shape,
 171                                                          filters_arg,
 172                                                          data_arg,
 173                                                          node->get_strides(),
 174                                                          node->get_dilations(),
 175                                                          node->get_pads_begin(),
 176                                                          node->get_pads_end(),
 177                                                          Strides(num_spatial_dims, 1));
 178         replace_node(node, replacement_node);
 179         return replacement_node;
 180     }
 181
 182     shared_ptr<Node> op_cast(shared_ptr<op::v1::Divide> node)
 183     {
 184         const auto input_arg0 = node->input_value(0);
 185         const auto input_arg1 = node->input_value(1);
 186         const auto autob = node->get_autob();
 187         const bool pydiv = node->is_pythondiv();
 188         auto replacement_node = make_shared<op::v0::Divide>(input_arg0, input_arg1, pydiv, autob);
 189         replace_node(node, replacement_node);
 190         return replacement_node;
 191     }
 192
 193     shared_ptr<Node> op_cast(shared_ptr<op::v1::Reshape> node)
 194     {
 195         shared_ptr<Node> replacement_node;
 196
 197         const auto target_shape_input = node->input_value(1).get_node_shared_ptr();
 198         const auto input_rank = node->get_input_partial_shape(0).rank();
 199         if (op::is_constant(target_shape_input) && node->get_output_partial_shape(0).is_static() &&
 200             input_rank.is_static())
 201         {
 202             const auto output_shape = node->get_output_shape(0);
 203             replacement_node = make_shared<op::Reshape>(
 204                 node->input_value(0), get_default_order(input_rank.get_length()), output_shape);
 205         }
 206         else
 207         {
 208             NGRAPH_CHECK(replacement_node, "Unable to convert Reshape:v1 with dynamic shape.");
 209         }
 210
 211         replace_node(node, replacement_node);
 212         return replacement_node;
 213     }
 214
 215     shared_ptr<Node> op_cast(shared_ptr<op::v1::Equal> node)
 216     {
 217         return op_cast_binary_elementwise_node<op::v0::Equal, op::v1::Equal>(node);
 218     }
 219
 220     shared_ptr<Node> op_cast(shared_ptr<op::v1::Greater> node)
 221     {
 222         return op_cast_binary_elementwise_node<op::v0::Greater, op::v1::Greater>(node);
 223     }
 224
 225     shared_ptr<Node> op_cast(shared_ptr<op::v1::GreaterEqual> node)
 226     {
 227         return op_cast_binary_elementwise_node<op::v0::GreaterEq, op::v1::GreaterEqual>(node);
 228     }
 229
 230     shared_ptr<Node> op_cast(shared_ptr<op::v1::GroupConvolution> node)
 231     {
 232         const auto data_arg = node->input_value(0);
 233         const auto filters_arg = node->input_value(1);
 234         const auto strides = node->get_strides();
 235         const size_t num_spatial_dims = strides.size();
 236         auto replacement_node = make_shared<op::v0::GroupConvolution>(data_arg,
 237                                                                       filters_arg,
 238                                                                       node->get_strides(),
 239                                                                       node->get_dilations(),
 240                                                                       node->get_pads_begin(),
 241                                                                       node->get_pads_end(),
 242                                                                       Strides(num_spatial_dims, 1),
 243                                                                       node->get_auto_pad());
 244         replace_node(node, replacement_node);
 245         return replacement_node;
 246     }
 247
 248     shared_ptr<Node> op_cast(shared_ptr<op::v1::GroupConvolutionBackpropData> node)
 249     {
 250         const auto data_arg = node->input_value(0);
 251         const auto filters_arg = node->input_value(1);
 252
 253         NGRAPH_CHECK(data_arg.get_partial_shape().is_static(),
 254                      "Unable to convert GroupConvolutionBackpropData:1 to "
 255                      "GroupConvolutionBackpropData:0 with dynamic data shape. Node: ",
 256                      *node);
 257
 258         NGRAPH_CHECK(filters_arg.get_partial_shape().is_static(),
 259                      "Unable to convert GroupConvolutionBackpropData:1 to "
 260                      "GroupConvolutionBackpropData:0 with dynamic filters shape. Node: ",
 261                      *node);
 262
 263         auto filters_shape = filters_arg.get_shape();
 264         const size_t groups = filters_shape.at(0);
 265
 266         const PartialShape output_pshape{node->get_output_partial_shape(0)};
 267         NGRAPH_CHECK(output_pshape.is_static(),
 268                      "Unable to convert GroupConvolutionBackpropData:v1 to "
 269                      "GroupConvolutionBackpropData:v0 "
 270                      "if output_shape is dynamic. Node: ",
 271                      *node);
 272         Shape output_shape = output_pshape.to_shape();
 273
 274         // Convert filters data layout from [GROUPS, C_INPUT, C_OUTPUT, K_D, ..., K_1]
 275         // into [C x M/group x k1 x k2 x ... x kn]
 276         filters_shape.erase(filters_shape.begin());
 277         filters_shape[0] *= groups;
 278
 279         auto reshaped_filters = builder::opset1::reshape(node->input_value(1), filters_shape);
 280
 281         auto replacement_node = make_shared<op::v0::GroupConvolutionBackpropData>(
 282             op::Constant::create(data_arg.get_element_type(), output_shape, {0}),
 283             reshaped_filters,
 284             data_arg,
 285             node->get_strides(),
 286             node->get_dilations(),
 287             node->get_pads_begin(),
 288             node->get_pads_end(),
 289             groups);
 290         replace_node(node, replacement_node);
 291         return replacement_node;
 292     }
 293
 294     shared_ptr<Node> op_cast(shared_ptr<op::v1::Less> node)
 295     {
 296         return op_cast_binary_elementwise_node<op::v0::Less, op::v1::Less>(node);
 297     }
 298
 299     shared_ptr<Node> op_cast(shared_ptr<op::v1::LessEqual> node)
 300     {
 301         return op_cast_binary_elementwise_node<op::v0::LessEq, op::v1::LessEqual>(node);
 302     }
 303
 304     shared_ptr<Node> op_cast(shared_ptr<op::v1::LogicalNot> node)
 305     {
 306         auto replacement_node = make_shared<op::v0::Not>(node->input_value(0));
 307         replace_node(node, replacement_node);
 308         return replacement_node;
 309     }
 310
 311     shared_ptr<Node> op_cast(shared_ptr<op::v1::LogicalOr> node)
 312     {
 313         return op_cast_binary_elementwise_node<op::v0::Or, op::v1::LogicalOr>(node);
 314     }
 315
 316     shared_ptr<Node> op_cast(shared_ptr<op::v1::LogicalXor> node)
 317     {
 318         return op_cast_binary_elementwise_node<op::v0::Xor, op::v1::LogicalXor>(node);
 319     }
 320
 321     shared_ptr<Node> op_cast(shared_ptr<op::v1::Maximum> node)
 322     {
 323         return op_cast_binary_elementwise_node<op::v0::Maximum, op::v1::Maximum>(node);
 324     }
 325
 326     shared_ptr<Node> op_cast(shared_ptr<op::v1::Minimum> node)
 327     {
 328         return op_cast_binary_elementwise_node<op::v0::Minimum, op::v1::Minimum>(node);
 329     }
 330
 331     shared_ptr<Node> op_cast(shared_ptr<op::v1::Multiply> node)
 332     {
 333         return op_cast_binary_elementwise_node<op::v0::Multiply, op::v1::Multiply>(node);
 334     }
 335
 336     shared_ptr<Node> op_cast(shared_ptr<op::v1::NotEqual> node)
 337     {
 338         return op_cast_binary_elementwise_node<op::v0::NotEqual, op::v1::NotEqual>(node);
 339     }
 340
 341     shared_ptr<Node> op_cast(shared_ptr<op::v1::Power> node)
 342     {
 343         return op_cast_binary_elementwise_node<op::v0::Power, op::v1::Power>(node);
 344     }
 345
 346     shared_ptr<Node> op_cast(shared_ptr<op::v1::ReduceMax> node)
 347     {
 348         auto replacement_node = op_cast_reduction_node<op::v0::Max, op::v1::ReduceMax>(node);
 349         replace_node(node, replacement_node);
 350         return replacement_node;
 351     }
 352
 353     shared_ptr<Node> op_cast(shared_ptr<op::v1::ReduceMean> node)
 354     {
 355         // ReduceMean = Sum / Count
 356         auto sum_node = op_cast_reduction_node<op::v0::Sum, op::v1::ReduceMean>(node);
 357
 358         // Count = Sum(Constant(1, shape=data.shape))
 359         const auto data = node->input_value(0);
 360         const auto axes = node->input_value(1);
 361         const auto const_node =
 362             op::v0::Constant::create(data.get_element_type(), data.get_shape(), {1});
 363         std::shared_ptr<Node> count_node = std::make_shared<op::v0::Sum>(const_node, axes);
 364
 365         // Support keep_dims attribute
 366         if (node->get_keep_dims())
 367         {
 368             // In order to keep the original dimensions we need to reshape the Count node
 369             // before we use it in Divide with NUMPY broadcast
 370             auto output_shape = count_node->get_shape();
 371             auto reshaped_output_shape = output_shape;
 372             for (const auto& axis : node->get_reduction_axes())
 373             {
 374                 reshaped_output_shape.insert(reshaped_output_shape.begin() + axis, 1);
 375             }
 376             count_node = make_shared<op::Reshape>(
 377                 count_node->output(0), get_default_order(output_shape), reshaped_output_shape);
 378         }
 379
 380         const auto replacement_node =
 381             std::make_shared<op::v0::Divide>(sum_node, count_node, op::AutoBroadcastSpec::NUMPY);
 382         replace_node(node, replacement_node);
 383         return replacement_node;
 384     }
 385
 386     shared_ptr<Node> op_cast(shared_ptr<op::v1::ReduceMin> node)
 387     {
 388         auto replacement_node = op_cast_reduction_node<op::v0::Min, op::v1::ReduceMin>(node);
 389         replace_node(node, replacement_node);
 390         return replacement_node;
 391     }
 392
 393     shared_ptr<Node> op_cast(shared_ptr<op::v1::ReduceSum> node)
 394     {
 395         auto replacement_node = op_cast_reduction_node<op::v0::Sum, op::v1::ReduceSum>(node);
 396         replace_node(node, replacement_node);
 397         return replacement_node;
 398     }
 399
 400     shared_ptr<Node> op_cast(shared_ptr<op::v1::Reverse> node)
 401     {
 402         auto axes_node = node->input_value(1).get_node_shared_ptr();
 403         NGRAPH_CHECK(op::is_constant(axes_node),
 404                      "Unable to convert Reverse:v1 to Reverse:v0 "
 405                      "if reduction axes are not constant. Node: ",
 406                      *node);
 407         const auto axes_node_const = as_type_ptr<op::Constant>(axes_node);
 408         AxisSet axes{};
 409         if (node->get_mode() == op::v1::Reverse::Mode::INDEX)
 410         {
 411             axes = axes_node_const->get_axis_vector_val();
 412         }
 413         else // Mode::MASK
 414         {
 415             auto axes_mask = axes_node_const->get_vector<bool>();
 416             for (size_t i = 0; i < axes_mask.size(); ++i)
 417             {
 418                 if (axes_mask[i])
 419                 {
 420                     axes.emplace(i);
 421                 }
 422             }
 423         }
 424         auto replacement_node = make_shared<op::v0::Reverse>(node->input_value(0), axes);
 425
 426         replace_node(node, replacement_node);
 427         return replacement_node;
 428     }
 429
 430     shared_ptr<Node> op_cast(shared_ptr<op::v1::Select> node)
 431     {
 432         ngraph::pass::ImplicitBroadcastElimination().run_on_node(node);
 433         auto replacement_node = make_shared<op::v0::Select>(
 434             node->input_value(0), node->input_value(1), node->input_value(2));
 435         replace_node(node, replacement_node);
 436         return replacement_node;
 437     }
 438
 439     shared_ptr<Node> op_cast(shared_ptr<op::v1::StridedSlice> node)
 440     {
 441         auto convert_mask_to_axes = [](const std::vector<int64_t>& mask) {
 442             AxisSet axes{};
 443             for (auto i = 0; i < mask.size(); ++i)
 444             {
 445                 if (mask[i] == 1)
 446                 {
 447                     axes.emplace(i);
 448                 }
 449             }
 450             return axes;
 451         };
 452
 453         const auto input_data = node->input_value(0);
 454         const auto input_data_pshape = input_data.get_partial_shape();
 455
 456         NGRAPH_CHECK(input_data_pshape.is_static(),
 457                      "Unable to convert StridedSlice:v1 to Slice:v0 "
 458                      "if input rank is not static. Node: ",
 459                      *node);
 460
 461         const auto begin_const =
 462             as_type_ptr<op::Constant>(node->input_value(1).get_node_shared_ptr());
 463         const auto end_const =
 464             as_type_ptr<op::Constant>(node->input_value(2).get_node_shared_ptr());
 465         const auto strides = as_type_ptr<op::Constant>(node->input_value(3).get_node_shared_ptr());
 466
 467         NGRAPH_CHECK(begin_const && end_const && strides,
 468                      "Unable to convert StridedSlice:v1 to Slice:v0 "
 469                      "if begin, end or strides are not constant. Node: ",
 470                      *node);
 471
 472         SlicePlan p = make_slice_plan(input_data_pshape.to_shape(),
 473                                       begin_const->get_vector<int64_t>(),
 474                                       end_const->get_vector<int64_t>(),
 475                                       strides->get_vector<int64_t>(),
 476                                       convert_mask_to_axes(node->get_begin_mask()),
 477                                       convert_mask_to_axes(node->get_end_mask()),
 478                                       convert_mask_to_axes(node->get_new_axis_mask()),
 479                                       convert_mask_to_axes(node->get_shrink_axis_mask()),
 480                                       convert_mask_to_axes(node->get_ellipsis_mask()));
 481
 482         shared_ptr<Node> replacement_node =
 483             make_shared<op::v0::Slice>(input_data,
 484                                        Coordinate(p.begins.begin(), p.begins.end()),
 485                                        Coordinate(p.ends.begin(), p.ends.end()),
 486                                        Strides(p.strides.begin(), p.strides.end()));
 487
 488         if (p.reshape_in_shape != p.reshape_out_shape)
 489         {
 490             replacement_node =
 491                 make_shared<op::Reshape>(replacement_node,
 492                                          ngraph::get_default_order(p.reshape_in_shape),
 493                                          p.reshape_out_shape);
 494         }
 495
 496         if (!p.reverse_axes.empty())
 497         {
 498             replacement_node = make_shared<op::Reverse>(replacement_node, p.reverse_axes);
 499         }
 500
 501         replace_node(node, replacement_node);
 502         return replacement_node;
 503     }
 504
 505     shared_ptr<Node> op_cast(shared_ptr<op::v1::Split> node)
 506     {
 507         const auto num_splits = node->get_num_splits();
 508
 509         auto replacement_node =
 510             make_shared<op::v0::Split>(node->input_value(0), node->input_value(1), num_splits);
 511
 512         replace_node(node, replacement_node);
 513         return replacement_node;
 514     }
 515
 516     shared_ptr<Node> op_cast(shared_ptr<op::v1::Subtract> node)
 517     {
 518         return op_cast_binary_elementwise_node<op::v0::Subtract, op::v1::Subtract>(node);
 519     }
 520
 521     shared_ptr<Node> op_cast(shared_ptr<op::v1::TopK> node)
 522     {
 523         const auto axis = node->get_axis();
 524         const auto sort_type = node->get_sort_type();
 525         const auto index_elem_type = node->get_index_element_type();
 526
 527         bool compute_max;
 528         switch (node->get_mode())
 529         {
 530         case op::v1::TopK::Mode::MAX: compute_max = true; break;
 531         case op::v1::TopK::Mode::MIN: compute_max = false; break;
 532         default: break;
 533         }
 534
 535         const auto arg_node = node->input_value(0);
 536         const auto k_node = node->input_value(1);
 537
 538         auto replacement_node = make_shared<op::v0::TopK>(
 539             arg_node, k_node, axis, index_elem_type, compute_max, sort_type);
 540
 541         // values output will be 0, indices 1
 542         vector<int64_t> output_order{1, 0};
 543         replace_node(node, replacement_node, output_order);
 544         return replacement_node;
 545     }
 546
 547     shared_ptr<Node> op_cast(shared_ptr<op::v1::Transpose> node)
 548     {
 549         const auto data = node->input_value(0);
 550
 551         const auto data_pshape = data.get_partial_shape();
 552         NGRAPH_CHECK(data_pshape.is_static(),
 553                      "Unable to convert Transpose:v1 to Reshape:v0 "
 554                      "if data shape is dynamic. Node: ",
 555                      *node);
 556         const auto data_shape = data_pshape.to_shape();
 557
 558         const auto order_node = node->input_value(1).get_node_shared_ptr();
 559         NGRAPH_CHECK(op::is_constant(order_node),
 560                      "Unable to convert Transpose:v1 to Reshape:v0 "
 561                      "if order node is not constant. Node: ",
 562                      *node);
 563         const auto order_const = as_type_ptr<op::Constant>(order_node);
 564
 565         auto order = order_const->get_axis_vector_val();
 566         Shape out_shape = data_shape;
 567         if (order.empty())
 568         {
 569             order.resize(out_shape.size());
 570             iota(begin(order), end(order), 0);
 571         }
 572         else
 573         {
 574             for (size_t i = 0; i < order.size(); ++i)
 575             {
 576                 out_shape[i] = data_shape.at(order.at(i));
 577             }
 578         }
 579
 580         auto replacement_node = make_shared<op::v0::Reshape>(data, order, out_shape);
 581         replace_node(node, replacement_node);
 582         return replacement_node;
 583     }
 584
 585     shared_ptr<Node> op_cast(shared_ptr<op::v1::VariadicSplit> node)
 586     {
 587         const auto split_lengths = node->input_value(2).get_node_shared_ptr();
 588
 589         NGRAPH_CHECK(op::is_constant(split_lengths),
 590                      "Unable to convert VariadicSplit:v1 to Split:v0 "
 591                      "if 'split_lengths' input is not constant. Node: ",
 592                      *node);
 593
 594         const auto splits = as_type_ptr<op::Constant>(split_lengths)->cast_vector<int64_t>();
 595         const std::vector<size_t> splits_unsigned{splits.begin(), splits.end()};
 596
 597         auto replacement_node =
 598             make_shared<op::v0::Split>(node->input_value(0), node->input_value(1), splits_unsigned);
 599
 600         replace_node(node, replacement_node);
 601         return replacement_node;
 602     }
 603
 604     using DispatchMap = map<NodeTypeInfo, std::function<bool(shared_ptr<Node> node)>>;
 605
 606     template <typename T>
 607     bool op_cast_thunk(shared_ptr<Node> node)
 608     {
 609         auto downgraded_node = op_cast(as_type_ptr<T>(node));
 610         if (downgraded_node)
 611         {
 612             if (ngraph::get_provenance_enabled())
 613             {
 614                 const std::string provenance_tag =
 615                     "<Opset0_Downgrade (v1 " + std::string(node->get_type_name()) + ")>";
 616                 downgraded_node->add_provenance_tags_above(node->input_values(), {provenance_tag});
 617             }
 618             return true;
 619         }
 620         return false;
 621     }
 622
 623     DispatchMap& get_dispatch_map()
 624     {
 625         static DispatchMap dispatch_map{
 626 #define NGRAPH_OP(NAME, NAMESPACE) {NAMESPACE::NAME::type_info, op_cast_thunk<NAMESPACE::NAME>},
 627 #include "ngraph/opsets/opset1_tbl.hpp"
 628 #undef NGRAPH_OP
 629         };
 630         return dispatch_map;
 631     }
 632 } // namespace opset0_downgrade
 633
 634 bool pass::Opset0Downgrade::run_on_node(shared_ptr<Node> node)
 635 {
 636     bool modified = false;
 637     auto& dispatch_map = opset0_downgrade::get_dispatch_map();
 638     auto it = dispatch_map.find(node->get_type_info());
 639     if (it != dispatch_map.end())
 640     {
 641         modified = it->second(node);
 642     }
 643     return modified;
 644 }