From 935550fa1078f2492573185513de5b0c7ee37f14 Mon Sep 17 00:00:00 2001
From: TolyaTalamanov <anatoliy.talamanov@intel.com>
Date: Thu, 16 Jan 2025 13:33:36 +0000
Subject: [PATCH] Not copy kvcache after last token generated

---
 src/plugins/intel_npu/src/plugin/npuw/llm_infer_request.cpp | 4 ++++
 1 file changed, 4 insertions(+)

diff --git a/src/plugins/intel_npu/src/plugin/npuw/llm_infer_request.cpp b/src/plugins/intel_npu/src/plugin/npuw/llm_infer_request.cpp
index 2e987036483e34..647a4b9f53142b 100644
--- a/src/plugins/intel_npu/src/plugin/npuw/llm_infer_request.cpp
+++ b/src/plugins/intel_npu/src/plugin/npuw/llm_infer_request.cpp
@@ -229,6 +229,10 @@ void ov::npuw::LLMInferRequest::infer_generate(ov::SoPtr<ov::ITensor> input_ids,
     m_logits = m_kvcache_request->get_tensor(m_kvcache_out_ports.at("logits"));
     kvcache_desc.num_stored_tokens += 1;
 
+    if (kvcache_desc.num_stored_tokens == kvcache_desc.total_size) {
+        return;
+    }
+
     LOG_DEBUG("Write KV-cache for the new token to the correct input position for next iteration.");
     const std::size_t kStartOutputKVCacheLayers = 1u;
     const auto& kvcache_compiled = m_kvcache_request->get_compiled_model();