diff --git a/src/plugins/intel_gpu/include/intel_gpu/graph/program.hpp b/src/plugins/intel_gpu/include/intel_gpu/graph/program.hpp
index b0f90bc1b82dee..7c5cba3a587290 100644
--- a/src/plugins/intel_gpu/include/intel_gpu/graph/program.hpp
+++ b/src/plugins/intel_gpu/include/intel_gpu/graph/program.hpp
@@ -288,9 +288,7 @@ struct program {
 
     void run_graph_compilation();
     void pre_optimize_graph(bool is_internal);
-#if 0 // TODO(Taylor)  temporally removed for multiple output PoC to reduce scope
     void post_optimize_graph(bool is_internal);
-#endif
     void cleanup();
     void transfer_memory_to_device();
 
diff --git a/src/plugins/intel_gpu/include/intel_gpu/primitives/concatenation.hpp b/src/plugins/intel_gpu/include/intel_gpu/primitives/concatenation.hpp
index c8c5f3c327c595..eb563ea6bd3e3c 100644
--- a/src/plugins/intel_gpu/include/intel_gpu/primitives/concatenation.hpp
+++ b/src/plugins/intel_gpu/include/intel_gpu/primitives/concatenation.hpp
@@ -55,7 +55,7 @@ struct concatenation : public primitive_base<concatenation> {
     /// @param axis Selected dimension for concatenation.
     concatenation(
         const primitive_id& id,
-        const input_info& input,
+        const std::vector<input_info>& input,
         const concatenation_axis axis,
         const primitive_id& ext_prim_id = "",
         const padding& output_padding = padding())
diff --git a/src/plugins/intel_gpu/include/intel_gpu/primitives/lstm_dynamic_input.hpp.backup b/src/plugins/intel_gpu/include/intel_gpu/primitives/lstm_dynamic_input.hpp
similarity index 80%
rename from src/plugins/intel_gpu/include/intel_gpu/primitives/lstm_dynamic_input.hpp.backup
rename to src/plugins/intel_gpu/include/intel_gpu/primitives/lstm_dynamic_input.hpp
index f993bcb8656b94..29b9110859626c 100644
--- a/src/plugins/intel_gpu/include/intel_gpu/primitives/lstm_dynamic_input.hpp.backup
+++ b/src/plugins/intel_gpu/include/intel_gpu/primitives/lstm_dynamic_input.hpp
@@ -35,13 +35,13 @@ struct lstm_dynamic_input : public primitive_base<lstm_dynamic_input> {
     /// @param recurrent Primitive id containing recurrent data.
     /// @param bias Primitive id containing bias data. Provide empty string if using lstm_dynamic without bias.
     lstm_dynamic_input(const primitive_id& id,
-                       const primitive_id& input,
+                       const input_info& input,
                        const primitive_id& dyn_length,
                        const primitive_id& weights,
                        const primitive_id& bias = "",
                        const primitive_id& ext_prim_id = "",
                        const padding& output_padding = padding())
-        : primitive_base(id, {input}, ext_prim_id, output_padding), dyn_length(dyn_length), weights(weights), bias(bias) {}
+        : primitive_base(id, {input}, ext_prim_id, {output_padding}), dyn_length(dyn_length), weights(weights), bias(bias) {}
 
     /// @brief Primitive id containing the dynamic sequence lengths.
     primitive_id dyn_length;
@@ -51,13 +51,13 @@ struct lstm_dynamic_input : public primitive_base<lstm_dynamic_input> {
     primitive_id bias;
 
 protected:
-    std::vector<std::reference_wrapper<const primitive_id>> get_dependencies() const override {
-        std::vector<std::reference_wrapper<const primitive_id>> ret;
-        ret.push_back(dyn_length);
-        ret.push_back(weights);
+    std::vector<std::pair<std::reference_wrapper<const primitive_id>, int>> get_dependencies() const override {
+        std::vector<std::pair<std::reference_wrapper<const primitive_id>, int>> ret;
+        ret.push_back({std::ref(dyn_length), 0});
+        ret.push_back({std::ref(weights), 0});
 
         if (!bias.empty()) {
-            ret.push_back(bias);
+            ret.push_back({std::ref(bias), 0});
         }
         return ret;
     }
diff --git a/src/plugins/intel_gpu/include/intel_gpu/primitives/split.hpp.backup b/src/plugins/intel_gpu/include/intel_gpu/primitives/split.hpp
similarity index 96%
rename from src/plugins/intel_gpu/include/intel_gpu/primitives/split.hpp.backup
rename to src/plugins/intel_gpu/include/intel_gpu/primitives/split.hpp
index cea5abedeee087..03252a9ade9705 100644
--- a/src/plugins/intel_gpu/include/intel_gpu/primitives/split.hpp.backup
+++ b/src/plugins/intel_gpu/include/intel_gpu/primitives/split.hpp
@@ -44,11 +44,11 @@ struct split : public primitive_base<split> {
     /// @param input Input primitive id.
     /// @param output_ids_offsets Pairs of output_ids and offsets
     split(const primitive_id& id,
-          const primitive_id& input,
+          const input_info& input,
           const std::vector<std::pair<primitive_id, tensor> >& output_ids_offsets,
           const primitive_id& ext_prim_id = "",
           const padding& output_padding = padding())
-        : primitive_base(id, {input}, ext_prim_id, output_padding),
+        : primitive_base(id, {input}, ext_prim_id, {output_padding}),
           output_offsets(extract_tensor_vector(output_ids_offsets)),
           output_ids(extract_primitive_vector(output_ids_offsets)) {}
 
diff --git a/src/plugins/intel_gpu/src/graph/graph_optimizer/calculate_prior_boxes.cpp.backup b/src/plugins/intel_gpu/src/graph/graph_optimizer/calculate_prior_boxes.cpp
similarity index 91%
rename from src/plugins/intel_gpu/src/graph/graph_optimizer/calculate_prior_boxes.cpp.backup
rename to src/plugins/intel_gpu/src/graph/graph_optimizer/calculate_prior_boxes.cpp
index 4eec60ac04431a..16c61eb42a3a9c 100644
--- a/src/plugins/intel_gpu/src/graph/graph_optimizer/calculate_prior_boxes.cpp.backup
+++ b/src/plugins/intel_gpu/src/graph/graph_optimizer/calculate_prior_boxes.cpp
@@ -26,7 +26,7 @@ void calculate_prior_boxes::run(program& p) {
 
         auto result = pb_node.get_result_buffer();
 
-        auto& data_node = p.get_or_create(std::make_shared<data>("_cldnn_tmp_" + pb_node.id() + "_result", result));
+        auto& data_node = p.get_or_create(std::make_shared<data>("_cldnn_tmp_" + pb_node.id() + "_result", std::vector<memory::ptr>{result}));
         p.replace(pb_node, data_node);
     }
 }
diff --git a/src/plugins/intel_gpu/src/graph/graph_optimizer/graph_initializations.cpp b/src/plugins/intel_gpu/src/graph/graph_optimizer/graph_initializations.cpp
index e44fa068cdfc15..5f854eaaa4dfcc 100644
--- a/src/plugins/intel_gpu/src/graph/graph_optimizer/graph_initializations.cpp
+++ b/src/plugins/intel_gpu/src/graph/graph_optimizer/graph_initializations.cpp
@@ -6,22 +6,19 @@
 
 #include "pass_manager.h"
 #include "program_node.h"
-#if 0 // TODO(taylor)
+
 #include "split_inst.h"
 #include "convolution_inst.h"
 #include "crop_inst.h"
 #include "lstm_inst.h"
 #include "reshape_inst.h"
 #include "resample_inst.h"
-#endif
 #include "permute_inst.h"
-#if 0 // TODO(taylor)
 #include "depth_to_space_inst.h"
 #include "lstm_dynamic_inst.h"
 #include "lstm_dynamic_input_inst.h"
 #include "lstm_dynamic_timeloop_inst.h"
 #include "mutable_data_inst.h"
-#endif
 #include "arg_max_min_inst.h"
 #include "kernel_selector_utils.h"
 
@@ -40,7 +37,7 @@ std::string get_id_string(size_t i) {
     ss << std::setw(5) << std::setfill('0') << i;
     return ss.str();
 }
-#if 0 // TODO(taylor)
+
 void graph_initializations::handle_split_node(program& p, split_node& node) {
     if (!node.get_users().empty()) {
         throw std::logic_error("Split layer cannot be used directly! Please use split output \"" + node.id() +
@@ -152,7 +149,7 @@ void graph_initializations::handle_lstm_node(program& p, lstm_node& node) {
                 primitive_id crop_id = input.id() + ":crop:" + get_id_string(sequence_element);
                 tensor crop_tensor{input_size.batch[0], 1, input_size.spatial[0], input_size.spatial[1]};
                 tensor offset_tensor{0, static_cast<tensor::value_type>(sequence_element), 0, 0};
-                auto input_crop = std::make_shared<crop>(crop_id, input.id(), crop_tensor, offset_tensor);
+                auto input_crop = std::make_shared<crop>(crop_id, input_info(input.id()), crop_tensor, offset_tensor);
                 auto& input_crop_node = p.get_or_create(input_crop);
 
                 // Add the crop nodes as user for input
@@ -217,10 +214,10 @@ void graph_initializations::handle_lstm_node(program& p, lstm_node& node) {
             // primitive_id lstm_gemm_input_id = node->get_dependency(input_idx).get_primitive()->id;
             // the line below requires an attention: get_org_primitive_id() might not be an actual id of a node
             // (see rename method) ToDO: ensure that get_org_primitive_id() is suitable here
-            primitive_id lstm_gemm_input_id = node.get_dependency(input_idx).get_org_primitive_id();
+            primitive_id lstm_gemm_input_id = node.get_dependency(input_idx).first->get_org_primitive_id();
 
             auto lstm_gemm_node = std::make_shared<lstm_gemm>(lstm_gemm_id,
-                                                              lstm_gemm_input_id,
+                                                              input_info(lstm_gemm_input_id),
                                                               weights_id,
                                                               recurrent_id,
                                                               bias_id,
@@ -229,7 +226,7 @@ void graph_initializations::handle_lstm_node(program& p, lstm_node& node) {
             auto& n1 = p.get_or_create(lstm_gemm_node);
 
             auto lstm_elt_node = std::make_shared<lstm_elt>(lstm_elt_id,
-                                                            lstm_gemm_id,
+                                                            input_info(lstm_gemm_id),
                                                             cell_id,
                                                             lstm_prim->clip,
                                                             lstm_prim->input_forget,
@@ -242,7 +239,7 @@ void graph_initializations::handle_lstm_node(program& p, lstm_node& node) {
             p.add_connection(n1, n2);
             // adding dependecy to lstm_gemm node
             // input
-            p.add_connection(node.get_dependency(input_idx), n1);
+            p.add_connection(*node.get_dependency(input_idx).first, n1);
             // adding weights and initial values to lstm_gemm
             p.add_connection(p.get_node(weights_id), n1);
             p.add_connection(p.get_node(recurrent_id), n1);
@@ -264,7 +261,7 @@ void graph_initializations::handle_lstm_node(program& p, lstm_node& node) {
             {
                 hidden_id = crop_id + ":hidden";
                 auto crop_hidden =
-                    std::make_shared<crop>(hidden_id, lstm_elt_id, hidden_size, tensor{0, 0, 0, 0});
+                    std::make_shared<crop>(hidden_id, input_info(lstm_elt_id), hidden_size, tensor{0, 0, 0, 0});
                 auto& n3 = p.get_or_create(crop_hidden);
                 // adding eltwise as dependency to hidden
                 p.add_connection(n2, n3);
@@ -284,7 +281,7 @@ void graph_initializations::handle_lstm_node(program& p, lstm_node& node) {
             // lstm_cell
             if (i < sequence_len - 1 || emit_last_cell) {
                 cell_id = crop_id + ":cell";
-                auto crop_cell = std::make_shared<crop>(cell_id, lstm_elt_id, hidden_size, tensor{0, 1, 0, 0});
+                auto crop_cell = std::make_shared<crop>(cell_id, input_info(lstm_elt_id), hidden_size, tensor{0, 1, 0, 0});
                 auto& n4 = p.get_or_create(crop_cell);
                 p.add_connection(n2, n4);
                 cell_list[i * directions + dir] = &n4;
@@ -296,9 +293,9 @@ void graph_initializations::handle_lstm_node(program& p, lstm_node& node) {
     }
     // if there is no next lstm, concatenation is created
     if (!has_lstm_children) {
-        std::vector<primitive_id> output_ids_offsets;
+        std::vector<input_info> output_ids_offsets;
         for (auto& e : output_map) {
-            output_ids_offsets.push_back(e.second.first);
+            output_ids_offsets.push_back(input_info(e.second.first));
         }
         primitive_id concatenation_id = node.id() + ":concat";
         auto concatenation_primitive =
@@ -319,7 +316,7 @@ void graph_initializations::handle_lstm_node(program& p, lstm_node& node) {
                                static_cast<int32_t>(concatenate_len),
                                hidden_size.spatial[0],
                                (int32_t)directions};
-            auto reshape_primitive = std::make_shared<reshape>(node.id() + ":reshape", concatenation_id, output_size);
+            auto reshape_primitive = std::make_shared<reshape>(node.id() + ":reshape", input_info(concatenation_id), output_size);
             auto& reshape_node = p.get_or_create(reshape_primitive);
             p.add_connection(concatenation_node, reshape_node);
             p.replace_all_usages(node, reshape_node);
@@ -345,12 +342,12 @@ void graph_initializations::handle_dynamic_lstm_node(program& p, lstm_dynamic_no
     // [1] Add lstm_dynamic_input
     auto lstm_dynamic_input_primitive =
         std::make_shared<lstm_dynamic_input>(node_id + suffix + "input",
-                                             input_id,
+                                             input_info(input_id),
                                              dyn_length_id,
                                              weights_id,
                                              bias_id,
                                              "",
-                                             node.get_primitive()->output_padding);
+                                             node.get_primitive()->output_paddings.at(0));
     auto& lstm_dynamic_input_node = p.get_or_create(lstm_dynamic_input_primitive);
     p.add_connection(node.input(), lstm_dynamic_input_node);  // connect real input to dlstm_input
     // connect other deps
@@ -367,7 +364,7 @@ void graph_initializations::handle_dynamic_lstm_node(program& p, lstm_dynamic_no
     auto last_cell_id = node.last_cell_state_id();
     auto lstm_dynamic_timeloop_primitive =
         std::make_shared<lstm_dynamic_timeloop>(node_id + suffix + "timeloop",
-                                                lstm_dynamic_input_node.id(),
+                                                input_info(lstm_dynamic_input_node.id()),
                                                 dyn_length_id,
                                                 recurrent_id,
                                                 last_hidden_id,
@@ -377,7 +374,7 @@ void graph_initializations::handle_dynamic_lstm_node(program& p, lstm_dynamic_no
                                                 node.clip(),
                                                 node.input_forget(),
                                                 "",
-                                                lstm_dynamic_input_primitive->output_padding);
+                                                lstm_dynamic_input_primitive->output_paddings.at(0));
     auto& lstm_dynamic_timeloop_node = p.get_or_create(lstm_dynamic_timeloop_primitive);
     p.add_connection(lstm_dynamic_input_node, lstm_dynamic_timeloop_node);  // connect dlstm_input to dlstm_timeloop
     // connect other deps
@@ -406,7 +403,7 @@ void graph_initializations::handle_dynamic_lstm_node(program& p, lstm_dynamic_no
 
     // we dont have to set output since it will be done in next graph_opts step
 }
-#endif
+
 void graph_initializations::set_outputs(program& p) {
     auto outputs_option = p.get_options().get<build_option_type::outputs>();
     if (!outputs_option->outputs.empty()) {
@@ -425,7 +422,6 @@ void graph_initializations::set_outputs(program& p) {
 }
 
 void graph_initializations::run(program& p) {
-#if 0 // TODO(taylor)
     auto itr = p.nodes_map.begin();
     while (itr != p.nodes_map.end()) {
         auto node_itr = itr++;
@@ -438,7 +434,6 @@ void graph_initializations::run(program& p) {
             handle_dynamic_lstm_node(p, node->as<lstm_dynamic>());
         }
     }
-#endif
     set_outputs(p);
     p.get_processing_order().calc_processing_order(p);
 }
diff --git a/src/plugins/intel_gpu/src/graph/graph_optimizer/post_input_reorder.cpp.backup b/src/plugins/intel_gpu/src/graph/graph_optimizer/post_input_reorder.cpp
similarity index 92%
rename from src/plugins/intel_gpu/src/graph/graph_optimizer/post_input_reorder.cpp.backup
rename to src/plugins/intel_gpu/src/graph/graph_optimizer/post_input_reorder.cpp
index d2ab615804db38..331e842698fb7f 100644
--- a/src/plugins/intel_gpu/src/graph/graph_optimizer/post_input_reorder.cpp.backup
+++ b/src/plugins/intel_gpu/src/graph/graph_optimizer/post_input_reorder.cpp
@@ -26,7 +26,10 @@ program_node& post_input_reorder::add_reorder(program& p,
     auto& new_reorder_node = p.get_or_create(new_reorder);
 
     // ToDo: add a method to program class which adds an intermediate node given a node and its user
-    auto it = std::find(usr->get_dependencies().begin(), usr->get_dependencies().end(), node);
+    auto it = std::find_if(usr->get_dependencies().begin(), usr->get_dependencies().end(),
+    [&](const std::pair<program_node*, int>& dep) {
+        return node == dep.first;
+    });
     if (it == usr->get_dependencies().end()) {
         throw std::runtime_error("Inconcistency in topology description: user of a node is not present among its dependecies.");
     }
@@ -52,7 +55,7 @@ void post_input_reorder::run(program& p) {
                 *static_cast<kernel_selector::fully_connected_params*>(fc_impl->_kernel_data.params.get());
 
             auto layout_format = from_data_layout(fc_params.inputs[0].GetLayout());
-            auto& input = node->get_dependencies()[0];
+            auto& input = node->get_dependencies()[0].first;
             auto input_layout = input->get_output_layout();
 
             if (input_layout.format != layout_format) {
diff --git a/src/plugins/intel_gpu/src/graph/graph_optimizer/post_optimize_weights.cpp.backup b/src/plugins/intel_gpu/src/graph/graph_optimizer/post_optimize_weights.cpp
similarity index 96%
rename from src/plugins/intel_gpu/src/graph/graph_optimizer/post_optimize_weights.cpp.backup
rename to src/plugins/intel_gpu/src/graph/graph_optimizer/post_optimize_weights.cpp
index d404d4028d57a3..dfab5b097b980e 100644
--- a/src/plugins/intel_gpu/src/graph/graph_optimizer/post_optimize_weights.cpp.backup
+++ b/src/plugins/intel_gpu/src/graph/graph_optimizer/post_optimize_weights.cpp
@@ -34,7 +34,7 @@ void post_optimize_weights::optimize_weights(T& node, program& p) {
     auto& weights_reorder_params = impl->_weights_reorder_params;
 
     for (auto i = offsets.weights_offset; i < offsets.bias_offset; i++) {
-        auto& weights_node = node.get_dependency(i);
+        auto& weights_node = *node.get_dependency(i).first;
         auto weights_layout = weights_node.get_output_layout();
 
         auto reorders = _rf.get_weights_reorder(weights_node.id(), weights_layout, weights_reorder_params);
@@ -43,7 +43,7 @@ void post_optimize_weights::optimize_weights(T& node, program& p) {
             // insert new generic_layer node to topology
             p.add_intermediate(reorder.first, node, i, !reorder.second);
             // set generic_layer's node output layout and implementation
-            auto& g_node = node.get_dependency(i);
+            auto& g_node = *node.get_dependency(i).first;
             g_node.get_output_layout(false);
 
             // Don't run impl selection to avoid double compilation of reorder kernels
diff --git a/src/plugins/intel_gpu/src/graph/graph_optimizer/update_loop_primitive_map.cpp.backup b/src/plugins/intel_gpu/src/graph/graph_optimizer/update_loop_primitive_map.cpp
similarity index 100%
rename from src/plugins/intel_gpu/src/graph/graph_optimizer/update_loop_primitive_map.cpp.backup
rename to src/plugins/intel_gpu/src/graph/graph_optimizer/update_loop_primitive_map.cpp
diff --git a/src/plugins/intel_gpu/src/graph/include/lstm_dynamic_input_inst.h.backup b/src/plugins/intel_gpu/src/graph/include/lstm_dynamic_input_inst.h
similarity index 89%
rename from src/plugins/intel_gpu/src/graph/include/lstm_dynamic_input_inst.h.backup
rename to src/plugins/intel_gpu/src/graph/include/lstm_dynamic_input_inst.h
index f92733f544c96d..d2d7f3756fb1e7 100644
--- a/src/plugins/intel_gpu/src/graph/include/lstm_dynamic_input_inst.h.backup
+++ b/src/plugins/intel_gpu/src/graph/include/lstm_dynamic_input_inst.h
@@ -19,13 +19,13 @@ struct typed_program_node<lstm_dynamic_input> : public typed_program_node_base<l
 public:
     typed_program_node(std::shared_ptr<primitive> prim, program& prog) : parent(prim, prog) {}
 
-    program_node& input() const { return get_dependency(0); }
-    program_node& dyn_length() const { return get_dependency(1); }
-    program_node& weights() const { return get_dependency(2); }
+    program_node& input() const { return *get_dependency(0).first; }
+    program_node& dyn_length() const { return *get_dependency(1).first; }
+    program_node& weights() const { return *get_dependency(2).first; }
 
     program_node& bias() const {
         CLDNN_ERROR_BOOL(id(), "Bias term", !bias_term(), "Trying to get non existing bias.");
-        return get_dependency(3);
+        return *get_dependency(3).first;
     }
 
     int32_t direction() const { return weights().get_output_layout().size.feature[0]; }
diff --git a/src/plugins/intel_gpu/src/graph/include/pass_manager.h b/src/plugins/intel_gpu/src/graph/include/pass_manager.h
index 94753749bb3908..8991b39d0d031c 100644
--- a/src/plugins/intel_gpu/src/graph/include/pass_manager.h
+++ b/src/plugins/intel_gpu/src/graph/include/pass_manager.h
@@ -6,14 +6,11 @@
 
 #include "intel_gpu/graph/program.hpp"
 #include "layout_optimizer.h"
-#if 0 // TODO(taylor)
 #include "split_inst.h"
 #include "lstm_inst.h"
 #include "lstm_dynamic_inst.h"
-#endif
 #include "quantize_inst.h"
 #include "eltwise_inst.h"
-#include "reorder_inst.h"
 #include "convolution_inst.h"
 #include "program_node.h"
 #include <string>
@@ -73,14 +70,14 @@ class add_reshape_to_primitives : public base_pass {
 private:
     void run(program& p) override;
 };
-#if 0 // TODO(taylor)
+
 class calculate_prior_boxes : public base_pass {
 public:
     calculate_prior_boxes() : base_pass("calculated_prior_boxes") {}
 private:
     void run(program& p) override;
 };
-#endif
+
 class compile_graph : public base_pass {
 public:
     compile_graph() : base_pass("compile_graph") {}
@@ -112,11 +109,9 @@ class graph_initializations : public base_pass {
 
 private:
     void run(program& p) override;
-#if 0 // TODO(taylor)
     void handle_split_node(program& p, split_node& node);
     void handle_lstm_node(program& p, lstm_node& node);
     void handle_dynamic_lstm_node(program& p, lstm_dynamic_node& node);
-#endif
     void set_outputs(program& p);
 };
 
@@ -237,7 +232,7 @@ class prepare_padding : public base_pass {
     void run(program& p) override;
     bool output_size_handling_enabled;
 };
-#if 0 // TODO(andrew)
+
 class post_input_reorder : public base_pass {
 public:
     post_input_reorder() : base_pass("post_input_reorder") {}
@@ -270,7 +265,7 @@ class post_optimize_weights : public base_pass {
     void optimize_weights(T& node, program& p);
     reorder_factory& _rf;
 };
-#endif
+
 class propagate_constants : public base_pass {
 public:
     propagate_constants() : base_pass("propagate_constants") {}
@@ -389,7 +384,7 @@ class oooq_memory_dependencies : public memory_dependency_pass {
     oooq_memory_dependencies() : memory_dependency_pass("oooq_memory_dependencies") {}
     void run(program& p) override;
 };
-
+#endif
 class update_loop_primitive_map : public base_pass {
 public:
     update_loop_primitive_map() : base_pass("update_loop_primitive_map") {}
@@ -397,7 +392,7 @@ class update_loop_primitive_map : public base_pass {
 private:
     void run(program& p) override;
 };
-#endif
+
 class add_onednn_optimization_attributes : public base_pass {
 public:
     add_onednn_optimization_attributes() : base_pass("add_onednn_optimization_attributes") {}
diff --git a/src/plugins/intel_gpu/src/graph/include/split_inst.h.backup b/src/plugins/intel_gpu/src/graph/include/split_inst.h
similarity index 93%
rename from src/plugins/intel_gpu/src/graph/include/split_inst.h.backup
rename to src/plugins/intel_gpu/src/graph/include/split_inst.h
index 13dbf8c9804868..438ff24eb945bb 100644
--- a/src/plugins/intel_gpu/src/graph/include/split_inst.h.backup
+++ b/src/plugins/intel_gpu/src/graph/include/split_inst.h
@@ -19,7 +19,7 @@ class typed_program_node<split> : public typed_program_node_base<split> {
 public:
     using parent::parent;
 
-    program_node& input() const { return get_dependency(0); }
+    program_node& input() const { return *get_dependency(0).first; }
 };
 
 using split_node = typed_program_node<split>;
diff --git a/src/plugins/intel_gpu/src/graph/lstm.cpp.backup b/src/plugins/intel_gpu/src/graph/lstm.cpp
similarity index 97%
rename from src/plugins/intel_gpu/src/graph/lstm.cpp.backup
rename to src/plugins/intel_gpu/src/graph/lstm.cpp
index fe90e8e24c69bd..b40c4d6c391092 100644
--- a/src/plugins/intel_gpu/src/graph/lstm.cpp.backup
+++ b/src/plugins/intel_gpu/src/graph/lstm.cpp
@@ -16,7 +16,7 @@ primitive_type_id lstm::type_id() {
 }
 
 layout lstm_inst::calc_output_layout(lstm_node const& node) {
-    assert(static_cast<bool>(node.get_primitive()->output_data_type) == false &&
+    assert(!node.get_primitive()->output_data_types.empty() &&
            "Output data type forcing is not supported for lstm_node!");
     auto input_layout = node.input().get_output_layout();
     auto hidden_layout = node.inital_hidden().get_output_layout();
diff --git a/src/plugins/intel_gpu/src/graph/lstm_dynamic_input.cpp.backup b/src/plugins/intel_gpu/src/graph/lstm_dynamic_input.cpp
similarity index 98%
rename from src/plugins/intel_gpu/src/graph/lstm_dynamic_input.cpp.backup
rename to src/plugins/intel_gpu/src/graph/lstm_dynamic_input.cpp
index a2f4dda3ed0ada..0ddda66953a99f 100644
--- a/src/plugins/intel_gpu/src/graph/lstm_dynamic_input.cpp.backup
+++ b/src/plugins/intel_gpu/src/graph/lstm_dynamic_input.cpp
@@ -19,7 +19,7 @@ primitive_type_id lstm_dynamic_input::type_id() {
 // weights_tensor: [b: 1, f: direction, x: input_size, y: 4 * hidden_size]
 // output_tensor:  [b: batch, f: max_sequence_length, x: 4 * hidden_size, y: direction]
 layout lstm_dynamic_input_inst::calc_output_layout(lstm_dynamic_input_node const& node) {
-    assert(static_cast<bool>(node.get_primitive()->output_data_type) == false &&
+    assert(!node.get_primitive()->output_data_types.empty() &&
            "Output data type forcing is not supported for lstm_dynamic_node!");
     auto input_layout = node.input().get_output_layout();
     auto weight_layout = node.weights().get_output_layout();
diff --git a/src/plugins/intel_gpu/src/graph/program.cpp b/src/plugins/intel_gpu/src/graph/program.cpp
index 4ec8b31e47c5b6..a9aec7e19c13cc 100644
--- a/src/plugins/intel_gpu/src/graph/program.cpp
+++ b/src/plugins/intel_gpu/src/graph/program.cpp
@@ -22,19 +22,15 @@
 
 #include "roi_pooling_inst.h"
 #include "reorg_yolo_inst.h"
-#if 0 // TODO(taylor)
 #include "eltwise_inst.h"
-#endif
 #include "softmax_inst.h"
 #include "permute_inst.h"
 #include "custom_gpu_primitive_inst.h"
 #include "binary_convolution_inst.h"
 #include "resample_inst.h"
 #include "reshape_inst.h"
-#if 0 // TODO(andrew)
 #include "quantize_inst.h"
 #include "activation_inst.h"
-#endif
 #include "scale_inst.h"
 #include "depth_to_space_inst.h"
 #include "convolution_inst.h"
@@ -43,8 +39,6 @@
 #include "data_inst.h"
 #include "deconvolution_inst.h"
 #include "detection_output_inst.h"
-#include "fully_connected_inst.h"
-#include "gather_inst.h"
 #include "input_layout_inst.h"
 #include "shuffle_channels_inst.h"
 #include "arg_max_min_inst.h"
@@ -52,25 +46,22 @@
 #include "lstm_elt_inst.h"
 #include "lstm_gemm_inst.h"
 #include "mutable_data_inst.h"
-#include "normalize_inst.h"
 #include "pooling_inst.h"
 #include "border_inst.h"
 #include "primitive_inst.h"
 #include "prior_box_inst.h"
 #include "proposal_inst.h"
 #include "reorder_inst.h"
-#if 0 // TODO(andrew)
 #include "split_inst.h"
-#endif
 #include "mvn_inst.h"
-#if 0 // TODO(andrew)
 #include "gemm_inst.h"
-#endif
 #include "reduce_inst.h"
 #include "region_yolo_inst.h"
 #include "strided_slice_inst.h"
 #include "loop_inst.h"
-
+// TODO(Andrew): Will be removed after ocl register enabled
+#include "gather_inst.h"
+#include "normalize_inst.h"
 #include "to_string_utils.h"
 #include "runtime/cldnn_itt.hpp"
 #include "runtime/kernels_cache.hpp"
@@ -452,9 +443,7 @@ void program::build_program(bool is_internal) {
     init_graph();
     { pre_optimize_graph(is_internal); }
     run_graph_compilation();
-#if 0 // TODO(taylor)
     { post_optimize_graph(is_internal); }
-#endif
 
     GPU_DEBUG_GET_INSTANCE(debug_config);
 #ifdef GPU_DEBUG_CONFIG
@@ -483,9 +472,8 @@ void program::build_program(bool is_internal) {
 void program::init_graph() {
     OV_ITT_SCOPED_TASK(itt::domains::CLDNN, "ProgramImpl::InitGraph");
     apply_opt_pass<graph_initializations>();
-#if 0 // TODO(taylor)
+
     apply_opt_pass<calculate_prior_boxes>();
-#endif
 
     apply_opt_pass<mark_nodes>();
 }
@@ -573,7 +561,7 @@ void program::pre_optimize_graph(bool is_internal) {
     // add optimization attributes for onednn primitives
     apply_opt_pass<add_onednn_optimization_attributes>();
 }
-#if 0 // TODO(taylor)
+
 void program::post_optimize_graph(bool is_internal) {
     OV_ITT_SCOPED_TASK(itt::domains::CLDNN, "ProgramImpl::PostOptimizeGraph");
     // input reorder for fully connected if necessary
@@ -596,7 +584,6 @@ void program::post_optimize_graph(bool is_internal) {
     // update loop input/output primitive mappings
     apply_opt_pass<update_loop_primitive_map>();
 }
-#endif
 
 // mark if the node is constant assuming that all dependencies are marked properly
 void program::mark_if_constant(program_node& node) {
diff --git a/src/plugins/intel_gpu/src/graph/split.cpp.backup b/src/plugins/intel_gpu/src/graph/split.cpp
similarity index 95%
rename from src/plugins/intel_gpu/src/graph/split.cpp.backup
rename to src/plugins/intel_gpu/src/graph/split.cpp
index 2265b916f2a0ae..ce91ce9cb6d033 100644
--- a/src/plugins/intel_gpu/src/graph/split.cpp.backup
+++ b/src/plugins/intel_gpu/src/graph/split.cpp
@@ -16,12 +16,12 @@ primitive_type_id split::type_id() {
 }
 
 layout split_inst::calc_output_layout(split_node const& node) {
-    assert(static_cast<bool>(node.get_primitive()->output_data_type) == false &&
+    assert(!node.get_primitive()->output_data_types.empty() &&
            "Output data type forcing is not supported for split_node!");
     auto output_ids = node.get_primitive()->output_ids;
     auto output_offsets = node.get_primitive()->output_offsets;
     auto param_num = output_ids.size();
-    auto input_sizes = node.get_dependency(0).get_non_padded_output_layout().size;
+    auto input_sizes = node.get_dependency(0).first->get_non_padded_output_layout().size;
     tensor null_tensor { 0, 0, 0, 0 };
 
     // check if output_ids count equals output_offsets count