[ONNX] Handle const scalar during unsqueeze (#6323)

* handle scalar during unsqueeze

* info about reverting

* use get_constant_from_source

* evaluate unsqueeze in onnx slice

ngraph/frontend/onnx_import/src/op/slice.cpp

@@ -12,6 +12,7 @@
 #include "ngraph/node.hpp"
 #include "ngraph/op/constant.hpp"
 #include "ngraph/op/util/op_types.hpp"
+#include "ngraph/validation_util.hpp"
 #include "op/gather.hpp"
 #include "utils/common.hpp"
 
@@ -171,13 +172,12 @@ namespace ngraph
                     auto ends = inputs.at(2);
 
                     // Slice is calculated over all axes as default
-                    Output<ngraph::Node> axes;
+                    std::shared_ptr<default_opset::Constant> axes_const;
                     if (inputs.size() >= 4 && !is_null(inputs.at(3))) // axes input provided
                     {
-                        axes = inputs.at(3);
-                        CHECK_VALID_NODE(node,
-                                         ngraph::op::is_constant(axes.get_node()),
-                                         "Axes input must be constant");
+                        axes_const = ngraph::get_constant_from_source(inputs.at(3));
+                        CHECK_VALID_NODE(
+                            node, axes_const != nullptr, "Axes input must be constant");
                     }
                     else
                     {
@@ -186,14 +186,11 @@ namespace ngraph
                             data_rank.is_static(),
                             "Data rank must be static when axes input is not provided");
                         const size_t data_rank_value = data_rank.get_length();
-                        axes = default_opset::Constant::create(
+                        axes_const = default_opset::Constant::create(
                             element::i64,
                             {data_rank_value},
                             common::get_monotonic_range<int64_t>(data_rank_value));
                     }
-
-                    const auto axes_const =
-                        as_type_ptr<default_opset::Constant>(axes.get_node_shared_ptr());
                     auto raw_axes_vec = axes_const->cast_vector<int64_t>();
                     std::vector<uint64_t> axes_vec =
                         get_normalized_axes_vector(node, data_rank, raw_axes_vec);

ngraph/test/models/onnx/slice_const_axes_source.prototxt

@@ -0,0 +1,114 @@
+ir_version: 7
+producer_name: "backend-test"
+graph {
+  name: "test_slice_with_unsqueeze_axes"
+  initializer {
+    data_type: 7
+    int64_data: 1
+    name: "x"
+  }
+  initializer {
+    dims: 1
+    data_type: 7
+    int64_data: 1
+    name: "starts"
+  }
+  initializer {
+    dims: 1
+    data_type: 7
+    int64_data: 3
+    name: "ends"
+  }
+  initializer {
+    dims: 1
+    data_type: 7
+    int64_data: 1
+    name: "steps"
+  }
+  node {
+    input: "x"
+    output: "slice_axes"
+    op_type: "Unsqueeze"
+    attribute {
+      name: "axes"
+      ints: 0
+      type: INTS
+    }
+  }
+  input {
+    name: "data"
+    type {
+      tensor_type {
+        elem_type: 1
+        shape {
+          dim {
+            dim_value:2 
+          }
+          dim {
+            dim_value:4
+          }
+        }
+      }
+    }
+  }
+  input {
+    name: "starts"
+    type {
+      tensor_type {
+        elem_type: 7
+        shape {
+          dim {
+            dim_value: 1
+          }
+        }
+      }
+    }
+  }
+  input {
+    name: "ends"
+    type {
+      tensor_type {
+        elem_type: 7
+        shape {
+          dim {
+            dim_value: 1
+          }
+        }
+      }
+    }
+  }
+  input {
+  name: "steps"
+    type {
+      tensor_type {
+        elem_type: 7
+        shape {
+          dim {
+            dim_value: 1
+          }
+        }
+      }
+    }
+  }
+  node {
+    input: "data"
+    input: "starts"
+    input: "ends"
+    input: "slice_axes"
+    input: "steps"
+    output: "sliced"
+    name: "Slice"
+    op_type: "Slice"
+  }
+  output {
+    name: "sliced"
+    type {
+      tensor_type {
+        elem_type: 1
+      }
+    }
+  }
+}
+opset_import {
+  version: 10
+}

ngraph/test/onnx/onnx_import.in.cpp

@@ -4334,6 +4334,17 @@ NGRAPH_TEST(${BACKEND_NAME}, onnx_multiple_slices_last_layer)
     test_case.run();
 }
 
+NGRAPH_TEST(${BACKEND_NAME}, onnx_slice_const_axes_source)
+{
+    auto function = onnx_import::import_onnx_model(
+        file_util::path_join(SERIALIZED_ZOO, "onnx/slice_const_axes_source.prototxt"));
+
+    auto test_case = test::TestCase<TestEngine>(function);
+    test_case.add_input<float>(std::vector<float>{1.f, 2.f, 3.f, 4.f, 5.f, 6.f, 7.f, 8.f});
+    test_case.add_expected_output<float>(Shape{2, 2}, {2.f, 3.f, 6.f, 7.f});
+    test_case.run();
+}
+
 NGRAPH_TEST(${BACKEND_NAME}, onnx_softmax_crossentropy_loss_mean)
 {
     auto function = onnx_import::import_onnx_model(