Apply suggestions from code review

Co-authored-by: Ilya Lavrenov <ilya.lavrenov@intel.com>

Author: popovaan

src/cpp/src/model_runner.hpp

@@ -209,7 +209,7 @@ class ModelRunner {
                             sequence_group->get_prompt_ids()[position_id] :
                             sequence->get_generated_ids()[position_id - prompt_len];
                     } else if (sequence_group_type == SequenceGroupType::EMBEDDINGS) {
-                        auto generated_embeds = sequence->get_generated_ids_embeds();
+                        const auto& generated_embeds = sequence->get_generated_ids_embeds();
                         const float* src = position_id < prompt_len ? sequence_group->get_input_embeds()[position_id].data() :  generated_embeds[position_id - prompt_len].data();
                         std::copy_n(src, hidden_size, inputs_embeds_data + token_id * hidden_size);
                     } else {
@@ -340,7 +340,7 @@ class ModelRunner {
             size_t seq_group_id = scheduler_output.m_scheduled_sequence_groups_ids[i];
             SequenceGroup::CPtr sequence_group = sequence_groups[seq_group_id];
             for (auto seq: sequence_group->get_running_sequences()) {
-                auto generated_ids = seq->get_generated_ids();
+                const auto& generated_ids = seq->get_generated_ids();
                 for (size_t token_idx = seq->get_generated_ids_embeds().size(); token_idx < generated_ids.size(); token_idx++) {
                     generated_ids_data[pos] = generated_ids[token_idx];
                     pos++;

src/cpp/src/sequence_group.cpp

@@ -40,7 +40,7 @@ size_t Sequence::_make_hash(size_t content_length) {
             size = content.size() * sizeof(content[0]);
         }
         else if (sequence_group->get_sequence_group_type() == SequenceGroupType::EMBEDDINGS) {
-            const auto input_embeds = sequence_group->get_input_embeds();
+            const auto& input_embeds = sequence_group->get_input_embeds();
             const auto generated_embeds = m_generated_ids_embeds;
             OPENVINO_ASSERT(content_length <= input_embeds.size() + generated_embeds.size());
             std::vector<float> content_float;
@@ -49,7 +49,6 @@ size_t Sequence::_make_hash(size_t content_length) {
             if (block_start_idx < input_embeds.size()) {
                 for (size_t idx = block_start_idx; idx < std::min(input_embeds.size(), content_length); idx++) {
                     auto embed = _reduce_embedding(input_embeds[idx]);
-                    const char* embed_char = reinterpret_cast<const char*>(embed.data());
                     content_float.insert(content_float.end(), embed.begin(), embed.end());
                 }
             }

src/cpp/src/sequence_group.hpp

@@ -202,7 +202,7 @@ class Sequence {
         m_sequence_group = sequence_group;
     }
 
-    std::vector<std::vector<float>> get_generated_ids_embeds() const{
+    const std::vector<std::vector<float>>& get_generated_ids_embeds() const {
         OPENVINO_ASSERT(m_type == ov::genai::SequenceGroupType::EMBEDDINGS);
         return m_generated_ids_embeds;
     }