diff --git a/nncf/common/graph/graph.py b/nncf/common/graph/graph.py
index 08a1fd587bd..756f3d3718e 100644
--- a/nncf/common/graph/graph.py
+++ b/nncf/common/graph/graph.py
@@ -597,12 +597,22 @@ def get_graph_for_structure_analysis(self, extended: bool = False) -> nx.DiGraph
             attrs_edge = {}
             u = u.replace(__RESERVED_DOT_CHARACTER, __CHARACTER_REPLACE_TO)
             v = v.replace(__RESERVED_DOT_CHARACTER, __CHARACTER_REPLACE_TO)
+            label = {}
+            if edge[NNCFGraph.PARALLEL_INPUT_PORT_IDS_ATTR]:
+                label["parallel_input_port_ids"] = edge[NNCFGraph.PARALLEL_INPUT_PORT_IDS_ATTR]
+
             if extended:
                 if edge[NNCFGraph.DTYPE_EDGE_ATTR] is Dtype.INTEGER:
                     attrs_edge["style"] = "dashed"
                 else:
                     attrs_edge["style"] = "solid"
-                attrs_edge["label"] = edge[NNCFGraph.ACTIVATION_SHAPE_EDGE_ATTR]
+                label["shape"] = edge[NNCFGraph.ACTIVATION_SHAPE_EDGE_ATTR]
+
+            if label:
+                if "shape" in label and len(label) == 1:
+                    attrs_edge["label"] = label["shape"]
+                else:
+                    attrs_edge["label"] = ", ".join((f"{k}:{v}" for k, v in label.items()))
             out_graph.add_edge(u, v, **attrs_edge)
         return out_graph
 
diff --git a/nncf/quantization/algorithms/min_max/algorithm.py b/nncf/quantization/algorithms/min_max/algorithm.py
index 1b199409683..ddaa6755cd4 100644
--- a/nncf/quantization/algorithms/min_max/algorithm.py
+++ b/nncf/quantization/algorithms/min_max/algorithm.py
@@ -505,7 +505,10 @@ def _get_quantization_target_points(
         hw_patterns = PatternsManager.get_full_hw_pattern_graph(backend=backend, device=device, model_type=model_type)
 
         inference_nncf_graph = transform_to_inference_graph(
-            deepcopy(nncf_graph), self._backend_entity.shapeof_metatypes, self._backend_entity.read_variable_metatypes
+            deepcopy(nncf_graph),
+            self._backend_entity.shapeof_metatypes,
+            self._backend_entity.dropout_metatypes,
+            self._backend_entity.read_variable_metatypes,
         )
 
         quantizer_setup = self._get_quantizer_setup(nncf_graph, inference_nncf_graph, hw_patterns, ignored_patterns)
diff --git a/nncf/quantization/algorithms/min_max/backend.py b/nncf/quantization/algorithms/min_max/backend.py
index 254a1c949cf..0764885a945 100644
--- a/nncf/quantization/algorithms/min_max/backend.py
+++ b/nncf/quantization/algorithms/min_max/backend.py
@@ -51,23 +51,23 @@ def post_processing_metatypes(self) -> List[OperatorMetatype]:
 
     @property
     @abstractmethod
-    def shapeof_metatypes(self) -> List[OperatorMetatype]:
+    def conv_metatypes(self) -> List[OperatorMetatype]:
         """
-        Property for the backend-specific ShapeOf metatypes.
+        Property for the backend-specific Convolution metatypes.
         """
 
     @property
     @abstractmethod
-    def conv_metatypes(self) -> List[OperatorMetatype]:
+    def shapeof_metatypes(self) -> List[OperatorMetatype]:
         """
-        Property for the backend-specific Convolution metatypes.
+        Property for the backend-specific ShapeOf metatypes.
         """
 
     @property
     @abstractmethod
-    def overflow_fix_metatypes(self) -> List[OperatorMetatype]:
+    def dropout_metatypes(self) -> List[OperatorMetatype]:
         """
-        Property for the backend-specific metatypes for which overflow_fix is applicable.
+        Property for the backend-specific Dropout metatypes.
         """
 
     @property
@@ -77,6 +77,13 @@ def read_variable_metatypes(self) -> List[OperatorMetatype]:
         Property for the backend-specific metatypes that also can be interpreted as inputs (ReadValue).
         """
 
+    @property
+    @abstractmethod
+    def overflow_fix_metatypes(self) -> List[OperatorMetatype]:
+        """
+        Property for the backend-specific metatypes for which overflow_fix is applicable.
+        """
+
     @property
     @abstractmethod
     def add_metatypes(self) -> List[OperatorMetatype]:
diff --git a/nncf/quantization/algorithms/min_max/onnx_backend.py b/nncf/quantization/algorithms/min_max/onnx_backend.py
index d3c1e25d0ae..e202252bf59 100644
--- a/nncf/quantization/algorithms/min_max/onnx_backend.py
+++ b/nncf/quantization/algorithms/min_max/onnx_backend.py
@@ -56,10 +56,6 @@ def mat_mul_metatypes(self) -> List[OperatorMetatype]:
     def post_processing_metatypes(self) -> List[OperatorMetatype]:
         return [om.ONNXTopKMetatype, om.ONNXNonMaxSuppressionMetatype]
 
-    @property
-    def shapeof_metatypes(self) -> List[OperatorMetatype]:
-        return [om.ONNXShapeMetatype]
-
     @property
     def conv_metatypes(self) -> List[OperatorMetatype]:
         return [om.ONNXConvolutionMetatype]
@@ -68,10 +64,6 @@ def conv_metatypes(self) -> List[OperatorMetatype]:
     def overflow_fix_metatypes(self) -> List[OperatorMetatype]:
         return [om.ONNXConvolutionMetatype, om.ONNXConvolutionTransposeMetatype, *MATMUL_METATYPES]
 
-    @property
-    def read_variable_metatypes(self) -> List[OperatorMetatype]:
-        return []
-
     @property
     def add_metatypes(self) -> List[OperatorMetatype]:
         return [om.ONNXAddLayerMetatype]
@@ -80,6 +72,18 @@ def add_metatypes(self) -> List[OperatorMetatype]:
     def group_conv_metatypes(self) -> List[OperatorMetatype]:
         return self.conv_metatypes
 
+    @property
+    def shapeof_metatypes(self) -> List[OperatorMetatype]:
+        return [om.ONNXShapeMetatype]
+
+    @property
+    def dropout_metatypes(self) -> List[OperatorMetatype]:
+        return []
+
+    @property
+    def read_variable_metatypes(self) -> List[OperatorMetatype]:
+        return []
+
     @property
     def scales_unification_map(self) -> Dict[OperatorMetatype, OperatorMetatype]:
         return {om.ONNXConcatMetatype: self.overflow_fix_metatypes}
diff --git a/nncf/quantization/algorithms/min_max/openvino_backend.py b/nncf/quantization/algorithms/min_max/openvino_backend.py
index 4edba0c0a37..103969142c2 100644
--- a/nncf/quantization/algorithms/min_max/openvino_backend.py
+++ b/nncf/quantization/algorithms/min_max/openvino_backend.py
@@ -56,10 +56,6 @@ def mat_mul_metatypes(self) -> List[OperatorMetatype]:
     def post_processing_metatypes(self) -> List[OperatorMetatype]:
         return [om.OVTopKMetatype, om.OVNonMaxSuppressionMetatype]
 
-    @property
-    def shapeof_metatypes(self) -> List[OperatorMetatype]:
-        return [om.OVShapeOfMetatype]
-
     @property
     def conv_metatypes(self) -> List[OperatorMetatype]:
         return [om.OVConvolutionMetatype]
@@ -74,10 +70,6 @@ def overflow_fix_metatypes(self) -> List[OperatorMetatype]:
             om.OVMatMulMetatype,
         ]
 
-    @property
-    def read_variable_metatypes(self) -> List[OperatorMetatype]:
-        return [om.OVReadValueMetatype]
-
     @property
     def add_metatypes(self) -> List[OperatorMetatype]:
         return [om.OVAddMetatype]
@@ -86,6 +78,18 @@ def add_metatypes(self) -> List[OperatorMetatype]:
     def group_conv_metatypes(self) -> List[OperatorMetatype]:
         return [om.OVGroupConvolutionMetatype]
 
+    @property
+    def shapeof_metatypes(self) -> List[OperatorMetatype]:
+        return [om.OVShapeOfMetatype]
+
+    @property
+    def dropout_metatypes(self) -> List[OperatorMetatype]:
+        return []
+
+    @property
+    def read_variable_metatypes(self) -> List[OperatorMetatype]:
+        return [om.OVReadValueMetatype]
+
     @property
     def scales_unification_map(self) -> Dict[OperatorMetatype, OperatorMetatype]:
         return {om.OVConcatMetatype: self.overflow_fix_metatypes}
diff --git a/nncf/quantization/algorithms/min_max/torch_backend.py b/nncf/quantization/algorithms/min_max/torch_backend.py
index e4a765a2d59..cae2c4fa6c2 100644
--- a/nncf/quantization/algorithms/min_max/torch_backend.py
+++ b/nncf/quantization/algorithms/min_max/torch_backend.py
@@ -69,6 +69,14 @@ def post_processing_metatypes(self) -> List[OperatorMetatype]:
     def shapeof_metatypes(self) -> List[OperatorMetatype]:
         return []
 
+    @property
+    def dropout_metatypes(self) -> List[OperatorMetatype]:
+        return [om.PTDropoutMetatype]
+
+    @property
+    def read_variable_metatypes(self) -> List[OperatorMetatype]:
+        return []
+
     @property
     def conv_metatypes(self) -> List[OperatorMetatype]:
         return [om.PTModuleConv1dMetatype, om.PTModuleConv2dMetatype, om.PTModuleConv3dMetatype]
@@ -85,10 +93,6 @@ def overflow_fix_metatypes(self) -> List[OperatorMetatype]:
             om.PTModuleConvTranspose3dMetatype,
         ]
 
-    @property
-    def read_variable_metatypes(self) -> List[OperatorMetatype]:
-        return []
-
     @property
     def add_metatypes(self) -> List[OperatorMetatype]:
         return [om.PTAddMetatype]
@@ -307,6 +311,10 @@ def get_ignored_metatypes(model_type: ModelType, device: TargetDevice) -> List[O
                 om.PTDivMetatype,
                 om.PTMaxMetatype,
                 om.PTSqueezeMetatype,
+                om.PTLayerNormMetatype,
+                om.PTModuleLayerNormMetatype,
+                om.PTGroupNormMetatype,
+                om.PTModuleGroupNormMetatype,
             ]
             if device != TargetDevice.CPU_SPR:
                 types.append(om.PTMulMetatype)
diff --git a/nncf/quantization/passes.py b/nncf/quantization/passes.py
index e6af74c8271..881b7f0a8c9 100644
--- a/nncf/quantization/passes.py
+++ b/nncf/quantization/passes.py
@@ -23,20 +23,23 @@
 def transform_to_inference_graph(
     nncf_graph: NNCFGraph,
     shapeof_metatypes: List[OperatorMetatype],
+    dropout_metatypes: List[OperatorMetatype],
     read_variable_metatypes: Optional[List[OperatorMetatype]] = None,
 ) -> NNCFGraph:
     """
-    This method contains pipeline of the passes that uses to provide inference graph without constant flows.
+    This method contains inplace pipeline of the passes that uses to provide inference graph without constant flows.
 
     :param nncf_graph: NNCFGraph instance for the transformation.
     :param shapeof_metatypes: List of backend-specific ShapeOf metatypes.
+    :param dropout_metatypes: List of backend-specific Dropout metatypes.
     :param read_variable_metatypes: List of backend-specific metatypes
         that also can be interpreted as inputs (ReadValue).
     :return: NNCFGraph in the inference style.
     """
-    inference_nncf_graph = remove_shapeof_subgraphs(nncf_graph, shapeof_metatypes, read_variable_metatypes)
-    inference_nncf_graph = filter_constant_nodes(nncf_graph, read_variable_metatypes)
-    return inference_nncf_graph
+    remove_shapeof_subgraphs(nncf_graph, shapeof_metatypes, read_variable_metatypes)
+    remove_nodes_and_reconnect_graph(nncf_graph, dropout_metatypes)
+    filter_constant_nodes(nncf_graph, read_variable_metatypes)
+    return nncf_graph
 
 
 def remove_shapeof_subgraphs(
@@ -45,7 +48,7 @@ def remove_shapeof_subgraphs(
     read_variable_metatypes: Optional[List[OperatorMetatype]] = None,
 ) -> NNCFGraph:
     """
-    Removes the ShapeOf subgraphs from the provided NNCFGraph instance.
+    Removes the ShapeOf subgraphs from the provided NNCFGraph instance inplace.
 
     :param nncf_graph: NNCFGraph instance for the transformation.
     :param shapeof_metatypes: List of backend-specific ShapeOf metatypes.
@@ -88,11 +91,61 @@ def remove_shapeof_subgraphs(
     return nncf_graph
 
 
+def remove_nodes_and_reconnect_graph(
+    nncf_graph: NNCFGraph,
+    metatypes: List[OperatorMetatype],
+) -> NNCFGraph:
+    """
+    Removes nodes with metatypes specified by `metatypes` parameter from
+    the provided NNCFGraph instance and connects previous node of a matched node
+    with next nodes of a matched node inplace for each matched node.
+    Matched nodes should have only one input node and only one output port.
+
+    :param nncf_graph: NNCFGraph instance for the transformation.
+    :param metatypes: List of backend-specific metatypes.
+    :return: Resulting NNCFGraph.
+    """
+    if not metatypes:
+        return nncf_graph
+
+    nodes_to_drop = []
+    for node in nncf_graph.get_nodes_by_metatypes(metatypes):
+        if node.metatype in metatypes:
+            nodes_to_drop.append(node)
+
+            prev_nodes = nncf_graph.get_previous_nodes(node)
+            input_edges = nncf_graph.get_input_edges(node)
+            assert len(prev_nodes) == len(input_edges) == 1
+            prev_node = prev_nodes[0]
+            input_edge = input_edges[0]
+            assert not input_edge.parallel_input_port_ids
+
+            # nncf_graph.get_next_edges is not used to preserve
+            # parallel_input_port_ids
+            for output_node in nncf_graph.get_next_nodes(node):
+                output_edge = nncf_graph.get_edge(node, output_node)
+                # Connects previous node with all next nodes
+                # to keep NNCFGraph connected.
+                assert input_edge.dtype == output_edge.dtype
+                assert input_edge.tensor_shape == output_edge.tensor_shape
+                nncf_graph.add_edge_between_nncf_nodes(
+                    from_node_id=prev_node.node_id,
+                    to_node_id=output_edge.to_node.node_id,
+                    tensor_shape=input_edge.tensor_shape,
+                    input_port_id=output_edge.input_port_id,
+                    output_port_id=input_edge.output_port_id,
+                    dtype=input_edge.dtype,
+                    parallel_input_port_ids=output_edge.parallel_input_port_ids,
+                )
+    nncf_graph.remove_nodes_from(nodes_to_drop)
+    return nncf_graph
+
+
 def filter_constant_nodes(
     nncf_graph: NNCFGraph, read_variable_metatypes: Optional[List[OperatorMetatype]] = None
 ) -> NNCFGraph:
     """
-    Removes all Constant nodes from NNCFGraph, making it inference graph.
+    Removes all Constant nodes from NNCFGraph inplace, making it inference graph.
     The traversing starts from the input nodes and nodes with weights.
 
     :param nncf_graph: NNCFGraph instance for the transformation.
diff --git a/nncf/torch/quantization/ignored_patterns.py b/nncf/torch/quantization/ignored_patterns.py
index d4816cc482e..799a709a1fb 100644
--- a/nncf/torch/quantization/ignored_patterns.py
+++ b/nncf/torch/quantization/ignored_patterns.py
@@ -67,8 +67,15 @@ def _add_softmax_reshape_matmul(
 
 @PT_IGNORED_PATTERNS.register(IgnoredPatternNames.MULTIHEAD_ATTENTION_OUTPUT)
 def create_multihead_attention_output() -> GraphPattern:
-    matmul_aliases = ["linear", "addmm", "matmul", "bmm", "mm", "baddbmm"]
-    reshape_squeeze_aliases = ["reshape", "view", "flatten", "squeeze", "unsqueeze", "squeeze", "flatten", "unsqueeze"]
+    matmul_aliases = ["linear", "addmm", "matmul", "bmm", "mm", "baddbmm", "__matmul__"]
+    reshape_squeeze_aliases = [
+        "reshape",
+        "view",
+        "flatten",
+        "unsqueeze",
+        "squeeze",
+        "unbind",
+    ]
     gather_aliases = ["gather", "index_select", "where", "index_select", "__getitem__"]
     transpose_aliases = ["transpose", "permute", "transpose_"]
 
diff --git a/tests/common/data/reference_graphs/passes/dropout_synthetic_model_after.dot b/tests/common/data/reference_graphs/passes/dropout_synthetic_model_after.dot
new file mode 100644
index 00000000000..bb1b1c72b51
--- /dev/null
+++ b/tests/common/data/reference_graphs/passes/dropout_synthetic_model_after.dot
@@ -0,0 +1,19 @@
+strict digraph  {
+"0 /Input_1_0" [id=0, type=Input_1];
+"1 /Split_1_0" [id=1, type=Split_1];
+"5 /Output_1_0" [id=5, type=Output_1];
+"6 /Output_2_1_0" [id=6, type=Output_2_1];
+"7 /Output_2_2_0" [id=7, type=Output_2_2];
+"8 /Output_2_3_0" [id=8, type=Output_2_3];
+"9 /Output_3_0" [id=9, type=Output_3];
+"10 /Output_2_4_0" [id=10, type=output];
+"11 /Output_3_1_0" [id=11, type=output];
+"0 /Input_1_0" -> "1 /Split_1_0";
+"1 /Split_1_0" -> "5 /Output_1_0";
+"1 /Split_1_0" -> "6 /Output_2_1_0";
+"1 /Split_1_0" -> "7 /Output_2_2_0";
+"1 /Split_1_0" -> "8 /Output_2_3_0";
+"1 /Split_1_0" -> "9 /Output_3_0";
+"1 /Split_1_0" -> "10 /Output_2_4_0";
+"1 /Split_1_0" -> "11 /Output_3_1_0"  [label="parallel_input_port_ids:[2, 3, 4, 5, 6, 7, 8, 9]"];
+}
diff --git a/tests/common/data/reference_graphs/passes/dropout_synthetic_model_before.dot b/tests/common/data/reference_graphs/passes/dropout_synthetic_model_before.dot
new file mode 100644
index 00000000000..f8c29563d5b
--- /dev/null
+++ b/tests/common/data/reference_graphs/passes/dropout_synthetic_model_before.dot
@@ -0,0 +1,25 @@
+strict digraph  {
+"0 /Input_1_0" [id=0, type=Input_1];
+"1 /Split_1_0" [id=1, type=Split_1];
+"2 /Dropout_1_0" [id=2, type=Dropout_1];
+"3 /Dropout_2_0" [id=3, type=Dropout_2];
+"4 /Dropout_3_0" [id=4, type=Dropout_3];
+"5 /Output_1_0" [id=5, type=Output_1];
+"6 /Output_2_1_0" [id=6, type=Output_2_1];
+"7 /Output_2_2_0" [id=7, type=Output_2_2];
+"8 /Output_2_3_0" [id=8, type=Output_2_3];
+"9 /Output_3_0" [id=9, type=Output_3];
+"10 /Output_2_4_0" [id=10, type=output];
+"11 /Output_3_1_0" [id=11, type=output];
+"0 /Input_1_0" -> "1 /Split_1_0";
+"1 /Split_1_0" -> "2 /Dropout_1_0";
+"1 /Split_1_0" -> "3 /Dropout_2_0";
+"1 /Split_1_0" -> "4 /Dropout_3_0";
+"2 /Dropout_1_0" -> "5 /Output_1_0";
+"3 /Dropout_2_0" -> "6 /Output_2_1_0";
+"3 /Dropout_2_0" -> "7 /Output_2_2_0";
+"3 /Dropout_2_0" -> "8 /Output_2_3_0";
+"3 /Dropout_2_0" -> "10 /Output_2_4_0";
+"4 /Dropout_3_0" -> "9 /Output_3_0";
+"4 /Dropout_3_0" -> "11 /Output_3_1_0"  [label="parallel_input_port_ids:[2, 3, 4, 5, 6, 7, 8, 9]"];
+}
diff --git a/tests/common/quantization/test_passes.py b/tests/common/quantization/test_passes.py
new file mode 100644
index 00000000000..c744c2297a4
--- /dev/null
+++ b/tests/common/quantization/test_passes.py
@@ -0,0 +1,54 @@
+# Copyright (c) 2023 Intel Corporation
+# Licensed under the Apache License, Version 2.0 (the "License");
+# you may not use this file except in compliance with the License.
+# You may obtain a copy of the License at
+#      http://www.apache.org/licenses/LICENSE-2.0
+# Unless required by applicable law or agreed to in writing, software
+# distributed under the License is distributed on an "AS IS" BASIS,
+# WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied.
+# See the License for the specific language governing permissions and
+# limitations under the License.
+
+from enum import Enum
+from pathlib import Path
+
+import pytest
+
+from nncf.quantization.passes import remove_nodes_and_reconnect_graph
+from tests.post_training.test_templates.models import NNCFGraphDropoutRemovingCase
+from tests.shared.nx_graph import compare_nx_graph_with_reference
+from tests.shared.paths import TEST_ROOT
+
+DATA_ROOT = TEST_ROOT / "common" / "data" / "reference_graphs"
+
+
+class TestModes(Enum):
+    VALID = "valid"
+    WRONG_TENSOR_SHAPE = "wrong_dropout_node"
+    WRONG_PARALLEL_EDGES = "wrong_parallel_edges"
+
+
+@pytest.mark.parametrize("mode", [TestModes.VALID, TestModes.WRONG_TENSOR_SHAPE, TestModes.WRONG_PARALLEL_EDGES])
+def test_remove_nodes_and_reconnect_graph(mode: TestModes):
+    def _check_graphs(dot_file_name, nncf_graph) -> None:
+        nx_graph = nncf_graph.get_graph_for_structure_analysis()
+        path_to_dot = DATA_ROOT / dot_file_name
+        compare_nx_graph_with_reference(nx_graph, path_to_dot, check_edge_attrs=True)
+
+    dot_reference_path_before = Path("passes") / "dropout_synthetic_model_before.dot"
+    dot_reference_path_after = Path("passes") / "dropout_synthetic_model_after.dot"
+    dropout_metatype = "DROPOUT_METATYPE"
+    kwargs = {}
+    if mode != TestModes.VALID:
+        kwargs.update({mode.value: True})
+
+    nncf_graph = NNCFGraphDropoutRemovingCase(dropout_metatype, **kwargs).nncf_graph
+
+    if mode != TestModes.VALID:
+        with pytest.raises(AssertionError):
+            remove_nodes_and_reconnect_graph(nncf_graph, [dropout_metatype])
+        return
+
+    _check_graphs(dot_reference_path_before, nncf_graph)
+    remove_nodes_and_reconnect_graph(nncf_graph, [dropout_metatype])
+    _check_graphs(dot_reference_path_after, nncf_graph)
diff --git a/tests/common/quantization/test_quantizer_propagation_solver.py b/tests/common/quantization/test_quantizer_propagation_solver.py
index 19937966b17..3193776aca1 100644
--- a/tests/common/quantization/test_quantizer_propagation_solver.py
+++ b/tests/common/quantization/test_quantizer_propagation_solver.py
@@ -1839,3 +1839,28 @@ def test_quantizers_are_not_set_up_for_integer_inputs(self, ip_graph_with_int_ed
         assert double_input_pq.current_location_node_key == InsertionPointGraph.get_pre_hook_node_key(
             "5 /E_0", input_port_id=1
         )
+
+
+def test_metatypes_to_ignore(mocker):
+    # pylint: disable=protected-access
+    NOT_IGNORED_METATYHPE = "not_ignored_metatype"
+    IGNORED_METATYPE = "target_metatype"
+
+    nncf_graph = NNCFGraph()
+    nodes = []
+    for node_name, node_metatype in zip("ABC", [NOT_IGNORED_METATYHPE, IGNORED_METATYPE, NOT_IGNORED_METATYHPE]):
+        nodes.append(nncf_graph.add_nncf_node(node_name, node_name, node_metatype=node_metatype))
+    for idx in range(1, len(nodes)):
+        nncf_graph.add_edge_between_nncf_nodes(
+            nodes[idx - 1].node_id, nodes[idx].node_id, [1, 1, 1, 1], 0, 0, Dtype.FLOAT
+        )
+    ip_graph = InsertionPointGraph(nncf_graph=nncf_graph, weight_modifiable_node_names=["A", "B", "C"])
+
+    solver = QuantizerPropagationSolver(
+        metatypes_to_ignore=[IGNORED_METATYPE],
+    )
+    solver._add_node_to_ignored = mocker.MagicMock()
+    solver.run_on_ip_graph(ip_graph)
+
+    solver._add_node_to_ignored.assert_called_once()
+    assert "1 B" in solver._add_node_to_ignored.call_args[0]
diff --git a/tests/post_training/reference_data.yaml b/tests/post_training/reference_data.yaml
index 250bfa3546b..2ea64d6cf1a 100644
--- a/tests/post_training/reference_data.yaml
+++ b/tests/post_training/reference_data.yaml
@@ -24,7 +24,7 @@ hf/hf-internal-testing/tiny-random-GPTNeoXForCausalLM_backend_OPTIMUM:
 
 # Timm
 timm/crossvit_9_240_backend_TORCH:
-    metric_value: 0.68136
+    metric_value: 0.689
     metric_value_fp32: 0.69966
 timm/crossvit_9_240_backend_ONNX:
     metric_value: 0.68906
@@ -56,7 +56,7 @@ timm/deit3_small_patch16_224_backend_OLD_TORCH:
     metric_value: 0.76514
     metric_value_fp32: 0.76974
 timm/deit3_small_patch16_224_backend_TORCH:
-    metric_value: 0.7621
+    metric_value: 0.76816
     metric_value_fp32: 0.76974
 timm/deit3_small_patch16_224_backend_ONNX:
     metric_value: 0.76806
@@ -276,7 +276,7 @@ timm/swin_base_patch4_window7_224_backend_OLD_TORCH:
     metric_value: 0.81376
     metric_value_fp32: 0.81462
 timm/swin_base_patch4_window7_224_backend_TORCH:
-    metric_value: 0.80696
+    metric_value: 0.8131
     metric_value_fp32: 0.81462
 timm/swin_base_patch4_window7_224_backend_ONNX:
     metric_value: 0.81294
diff --git a/tests/post_training/test_templates/models.py b/tests/post_training/test_templates/models.py
index 86d258066b4..24203fbeb21 100644
--- a/tests/post_training/test_templates/models.py
+++ b/tests/post_training/test_templates/models.py
@@ -10,6 +10,7 @@
 # limitations under the License.
 
 from nncf.common.graph import NNCFGraph
+from nncf.common.graph.layer_attributes import Dtype
 from nncf.common.graph.operator_metatypes import InputNoopMetatype
 from nncf.common.graph.operator_metatypes import OutputNoopMetatype
 from tests.common.quantization.metatypes import ConstantTestMetatype
@@ -218,3 +219,82 @@ def __init__(
             node_edges.extend([("Conv_2", "Output_1")])
         original_mock_graph = create_mock_graph(nodes, node_edges)
         self.nncf_graph = get_nncf_graph_from_mock_nx_graph(original_mock_graph, nncf_graph_cls)
+
+
+class NNCFGraphDropoutRemovingCase:
+    def __init__(
+        self,
+        dropout_metatype,
+        wrong_dropout_node: bool = False,
+        wrong_parallel_edges: bool = False,
+        nncf_graph_cls=NNCFGraph,
+    ):
+        nodes = [
+            NodeWithType("Input_1", InputNoopMetatype),
+            NodeWithType("Split_1", None),
+            NodeWithType(
+                "Dropout_1",
+                dropout_metatype,
+            ),
+            NodeWithType("Output_1", OutputNoopMetatype),
+            NodeWithType(
+                "Dropout_2",
+                dropout_metatype,
+            ),
+            NodeWithType("Output_2_1", OutputNoopMetatype),
+            NodeWithType("Output_2_2", OutputNoopMetatype),
+            NodeWithType("Output_2_3", OutputNoopMetatype),
+            NodeWithType(
+                "Dropout_3",
+                dropout_metatype,
+            ),
+            NodeWithType("Output_3", OutputNoopMetatype),
+        ]
+        node_edges = [
+            ("Input_1", "Split_1"),
+            ("Split_1", "Dropout_1"),
+            ("Dropout_1", "Output_1"),
+            ("Split_1", "Dropout_2"),
+            ("Dropout_2", "Output_2_1"),
+            ("Dropout_2", "Output_2_2"),
+            ("Dropout_2", "Output_2_3"),
+            ("Split_1", "Dropout_3"),
+            ("Dropout_3", "Output_3"),
+        ]
+        original_mock_graph = create_mock_graph(nodes, node_edges)
+        self.nncf_graph = get_nncf_graph_from_mock_nx_graph(original_mock_graph, nncf_graph_cls)
+
+        dropout_2 = self.nncf_graph.get_node_by_key("3 /Dropout_2_0")
+        output = self.nncf_graph.add_nncf_node("/Output_2_4_0", "output", OutputNoopMetatype)
+        tensor_shape = [1, 2, 1, 1] if wrong_dropout_node else [1, 1, 1, 1]
+        self.nncf_graph.add_edge_between_nncf_nodes(
+            dropout_2.node_id,
+            output.node_id,
+            tensor_shape=tensor_shape,
+            input_port_id=15,
+            output_port_id=1,
+            dtype=Dtype.FLOAT,
+        )
+
+        dropout_2 = self.nncf_graph.get_node_by_key("4 /Dropout_3_0")
+        output = self.nncf_graph.add_nncf_node("/Output_3_1_0", "output", OutputNoopMetatype)
+        self.nncf_graph.add_edge_between_nncf_nodes(
+            dropout_2.node_id,
+            output.node_id,
+            tensor_shape=tensor_shape,
+            input_port_id=1,
+            output_port_id=1,
+            dtype=Dtype.FLOAT,
+            parallel_input_port_ids=list(range(2, 10)),
+        )
+        if wrong_parallel_edges:
+            dropout_4 = self.nncf_graph.add_nncf_node("100 /dropout", "dropout", dropout_metatype)
+            self.nncf_graph.add_edge_between_nncf_nodes(
+                self.nncf_graph.get_node_by_key("0 /Input_1_0").node_id,
+                dropout_4.node_id,
+                tensor_shape=[1, 1, 1, 1],
+                input_port_id=0,
+                output_port_id=0,
+                dtype=Dtype.FLOAT,
+                parallel_input_port_ids=list(range(1, 10)),
+            )
diff --git a/tests/post_training/test_templates/test_ptq_params.py b/tests/post_training/test_templates/test_ptq_params.py
index 8de8ce10451..71a14d8ad76 100644
--- a/tests/post_training/test_templates/test_ptq_params.py
+++ b/tests/post_training/test_templates/test_ptq_params.py
@@ -165,6 +165,7 @@ def test_quantize_outputs(self, test_params, quantize_outputs):
         inference_nncf_graph = transform_to_inference_graph(
             deepcopy(nncf_graph),
             min_max_algo._backend_entity.shapeof_metatypes,
+            min_max_algo._backend_entity.dropout_metatypes,
             min_max_algo._backend_entity.read_variable_metatypes,
         )
         q_setup = min_max_algo._get_quantizer_setup(nncf_graph, inference_nncf_graph, hw_patterns, ignored_patterns)
@@ -189,6 +190,7 @@ def test_ignored_scopes(self, test_params, ignored_scopes_data):
         inference_nncf_graph = transform_to_inference_graph(
             deepcopy(nncf_graph),
             min_max_algo._backend_entity.shapeof_metatypes,
+            min_max_algo._backend_entity.dropout_metatypes,
             min_max_algo._backend_entity.read_variable_metatypes,
         )
         q_setup = min_max_algo._get_quantizer_setup(nncf_graph, inference_nncf_graph, hw_patterns, ignored_patterns)
@@ -213,6 +215,7 @@ def test_model_type_pass(self, test_params, model_type):
         inference_nncf_graph = transform_to_inference_graph(
             deepcopy(nncf_graph),
             min_max_algo._backend_entity.shapeof_metatypes,
+            min_max_algo._backend_entity.dropout_metatypes,
             min_max_algo._backend_entity.read_variable_metatypes,
         )
         q_setup = min_max_algo._get_quantizer_setup(nncf_graph, inference_nncf_graph, hw_patterns, ignored_patterns)
@@ -276,7 +279,7 @@ def test_quantization_points_overflow_fix(self, overflow_fix, affected_target_po
     @pytest.mark.parametrize("validate_scopes", (True, False))
     def test_validate_scope(self, test_params, validate_scopes):
         nncf_graph = test_params["test_model_type_pass"]["nncf_graph"]
-        inference_nncf_graph = transform_to_inference_graph(deepcopy(nncf_graph), [])
+        inference_nncf_graph = transform_to_inference_graph(deepcopy(nncf_graph), [], [])
         ignored_patterns = test_params["test_model_type_pass"]["ignored_patterns"]
         algo = MinMaxQuantization(
             ignored_scope=IgnoredScope(names=["some_node"], validate=validate_scopes),
diff --git a/tests/post_training/test_templates/test_quantizer_config.py b/tests/post_training/test_templates/test_quantizer_config.py
index 033b35377f5..48e5b585084 100644
--- a/tests/post_training/test_templates/test_quantizer_config.py
+++ b/tests/post_training/test_templates/test_quantizer_config.py
@@ -91,6 +91,7 @@ def test_default_quantizer_config(self, single_conv_nncf_graph):
         inference_nncf_graph = transform_to_inference_graph(
             deepcopy(nncf_graph),
             min_max_algo._backend_entity.shapeof_metatypes,
+            min_max_algo._backend_entity.dropout_metatypes,
             min_max_algo._backend_entity.read_variable_metatypes,
         )
         q_setup = min_max_algo._get_quantizer_setup(
@@ -144,6 +145,7 @@ def test_quantizer_config_from_ptq_params_for_CPU(
         inference_nncf_graph = transform_to_inference_graph(
             deepcopy(nncf_graph),
             min_max_algo._backend_entity.shapeof_metatypes,
+            min_max_algo._backend_entity.dropout_metatypes,
             min_max_algo._backend_entity.read_variable_metatypes,
         )
         if signed_weights is False or signed_activations in [True, False]:  # Incompatible with HW CPU config
@@ -185,6 +187,7 @@ def test_depthwise_conv_default_quantizer_config(self, depthwise_conv_nncf_graph
         inference_nncf_graph = transform_to_inference_graph(
             deepcopy(nncf_graph),
             min_max_algo._backend_entity.shapeof_metatypes,
+            min_max_algo._backend_entity.dropout_metatypes,
             min_max_algo._backend_entity.read_variable_metatypes,
         )
         q_setup = min_max_algo._get_quantizer_setup(