Move GRUCellFusion and SequenceFusion transformations before NopElimination (#14213)

src/common/transformations/src/transformations/common_optimizations/moc_transformations.cpp

@@ -146,6 +146,10 @@ bool ov::pass::MOCTransformations::run_on_model(const std::shared_ptr<ngraph::Fu
     REGISTER_PASS(manager, BroadcastElementwiseFusion)
     REGISTER_PASS(manager, PullThroughReduce)
 
+    // GRUCellFusion and SequenceFusion should be before NopElimination
+    REGISTER_PASS(manager, GRUCellFusion)
+    REGISTER_PASS(manager, SequenceFusion)
+
     auto transpose_sinking = manager.register_pass<ov::pass::GraphRewrite>();
     ADD_MATCHER(transpose_sinking, TransposeSinking)
 
@@ -153,7 +157,6 @@ bool ov::pass::MOCTransformations::run_on_model(const std::shared_ptr<ngraph::Fu
     // because it replaces pattern that may contain Transposes which must be optimized before
     // the transformation and it also inserts Transpose that can be optimized by TransposeSinking
     ADD_MATCHER(transpose_sinking, SplitSqueezeConcatFusion)
-
     auto eliminations = manager.register_pass<ov::pass::GraphRewrite>();
     ADD_MATCHER(eliminations, EliminateUnsqueezeGather)
     ADD_MATCHER(eliminations, NopElimination, m_use_shapes)
@@ -178,8 +181,6 @@ bool ov::pass::MOCTransformations::run_on_model(const std::shared_ptr<ngraph::Fu
     ADD_MATCHER(common_fusions, GeluFusion)
     ADD_MATCHER(common_fusions, LeakyReluFusion)
     ADD_MATCHER(common_fusions, RandomUniformFusion)
-    ADD_MATCHER(common_fusions, GRUCellFusion)
-    ADD_MATCHER(common_fusions, SequenceFusion)
     ADD_MATCHER(common_fusions, ConvertTensorIteratorToSequence)
     ADD_MATCHER(common_fusions, SplitConcatPairToInterpolateFusion, m_use_shapes)
     if (m_use_shapes) {