From 8ac5c225aa26a4a58a9d296842524be2f3e756a5 Mon Sep 17 00:00:00 2001
From: "Yu, Zijun" <zijun.yu@intel.com>
Date: Fri, 16 May 2025 10:12:22 +0800
Subject: [PATCH] FIX: set_max_token_len

---
 ggml/src/ggml-openvino/ggml-decoder.cpp | 5 +++--
 ggml/src/ggml-openvino/utils.cpp        | 2 +-
 2 files changed, 4 insertions(+), 3 deletions(-)

diff --git a/ggml/src/ggml-openvino/ggml-decoder.cpp b/ggml/src/ggml-openvino/ggml-decoder.cpp
index a8e1ad5556..e6474d6def 100644
--- a/ggml/src/ggml-openvino/ggml-decoder.cpp
+++ b/ggml/src/ggml-openvino/ggml-decoder.cpp
@@ -44,13 +44,14 @@ GgmlOvDecoder::GgmlOvDecoder(struct ggml_tensor* node, struct ggml_cgraph* cgrap
             dump_cgraph(m_cgraph);
         }
 
+        set_max_token_len();
+
         static bool weight_created = false;
         if (!getenv("GGML_OPENVINO_WEIGHT_AS_INPUT") && !weight_created) {
             add_weight_const_parallel(model_weights);
             weight_created = true;
         }
 
-        set_max_token_len();
         for (int node_n = 0; node_n < m_cgraph->n_nodes; node_n++) {
             auto* cur_node = m_cgraph->nodes[node_n];
             m_nodes.push_back(cur_node);
@@ -197,7 +198,7 @@ void GgmlOvDecoder::set_max_token_len() {
         auto* node = m_cgraph->nodes[i];
         if (std::string(node->name) == "k-0") {
             auto* cache_k = node->src[0];
-            m_max_token_len = cache_k->ne[0] / node->ne[0] / node->ne[1];
+            m_max_token_len = cache_k->ne[0] / node->ne[0] / node->ne[2];
             break;
         }
     }
diff --git a/ggml/src/ggml-openvino/utils.cpp b/ggml/src/ggml-openvino/utils.cpp
index 09bf0d0ac5..040ca1961e 100644
--- a/ggml/src/ggml-openvino/utils.cpp
+++ b/ggml/src/ggml-openvino/utils.cpp
@@ -209,4 +209,4 @@ void print_output_tensor_info(const std::string& name,
     default:
         break;
     }
-}
\ No newline at end of file
+}