Removed chat functionality as it is incomplete

openvinotoolkit · Dec 24, 2024 · bef8ca8 · bef8ca8
1 parent c52bd12
commit bef8ca8
Show file tree

Hide file tree

Showing 2 changed files with 3 additions and 25 deletions.
diff --git a/src/cpp/src/llm_pipeline_static.cpp b/src/cpp/src/llm_pipeline_static.cpp
@@ -745,16 +745,7 @@ DecodedResults StatefulLLMPipeline::generate(
         prompt = std::get<std::string>(inputs);
     }
 
-    ov::genai::TokenizedInputs tokenized_input;
-    if (m_is_chat_conversation) {
-        m_history.push_back({{"role", "user"}, {"content", prompt}});
-        constexpr bool add_generation_prompt = true;
-        prompt = m_tokenizer.apply_chat_template(m_history, add_generation_prompt);
-        // for chat ov::genai::add_special_tokens(false) is aligned with stateful pipeline and HF
-        tokenized_input = m_tokenizer.encode(prompt, ov::genai::add_special_tokens(false));
-    } else {
-        tokenized_input = m_tokenizer.encode(prompt);
-    }
+    ov::genai::TokenizedInputs tokenized_input = m_tokenizer.encode(prompt);
 
     auto encode_stop_time =  std::chrono::steady_clock::now();
     auto encoded_results = generate(tokenized_input, config, streamer);
@@ -763,11 +754,6 @@ DecodedResults StatefulLLMPipeline::generate(
     DecodedResults decoded_results = {m_tokenizer.decode(encoded_results.tokens), encoded_results.scores};
     auto decode_stop_time =  std::chrono::steady_clock::now();
 
-    if (m_is_chat_conversation) {
-        auto answer = decoded_results.texts[0];
-        m_history.push_back({{"role", "assistant"}, {"content", answer}});
-    }
-
     // generate_durations
     decoded_results.perf_metrics = encoded_results.perf_metrics;
     auto& raw_counters = decoded_results.perf_metrics.raw_metrics;
@@ -876,8 +862,6 @@ EncodedResults StatefulLLMPipeline::generate(
         if (last_token == config.eos_token_id && !config.ignore_eos) {
             break;
         }
-
-        // TODO: How to check that KV-Cache is full?
     }
 
     if (streamer_ptr) {
@@ -895,15 +879,11 @@ EncodedResults StatefulLLMPipeline::generate(
 }
 
 void StatefulLLMPipeline::start_chat(const std::string& system_message) {
-    if (!system_message.empty()) {
-        m_history.push_back({{"role", "system"}, {"content", system_message}});
-    }
-    m_is_chat_conversation = true;
+    // FIXME: Implement later
 };
 
 void StatefulLLMPipeline::finish_chat() {
-    m_is_chat_conversation = false;
-    m_history.clear();
+    // FIXME: Implement later
 };
 
 StatelessLLMPipeline::StatelessLLMPipeline(

diff --git a/src/cpp/src/llm_pipeline_static.hpp b/src/cpp/src/llm_pipeline_static.hpp
@@ -75,8 +75,6 @@ class StatefulLLMPipeline : public LLMPipelineImplBase {
 
 private:
     ov::InferRequest m_request;
-    bool m_is_chat_conversation = false;
-    ChatHistory m_history;
 };
 
 class StatelessLLMPipeline final : public LLMPipelineImplBase {