@@ -178,7 +178,6 @@ std::pair<std::vector<std::string>, std::unordered_set<std::string>> ExtractKVPa
178178// not_kv_inputs = ["input_ids", "attention_mask"]
179179std::pair<std::vector<std::string>, std::vector<std::string>> ExtractInputKVTensors (
180180 const std::shared_ptr<ov::Model>& model, const std::unordered_set<std::string>& kv_patterns) {
181-
182181 std::vector<std::string> key_value_input_names;
183182 std::vector<std::string> not_kv_inputs;
184183
@@ -258,11 +257,11 @@ void PatchStatefulDecoder(std::shared_ptr<ov::Model> model) {
258257 }
259258
260259 if (key_value_input_names.size () != key_value_output_names.size ()) {
261- ORT_THROW (" Found different sizes between key_value_input_names (" ,
262- key_value_input_names.size (),
263- " ) and key_value_output_names (" ,
264- key_value_output_names.size (),
265- " ). They couldn't be paired." );
260+ ORT_THROW (" Found different sizes between key_value_input_names (" ,
261+ key_value_input_names.size (),
262+ " ) and key_value_output_names (" ,
263+ key_value_output_names.size (),
264+ " ). They couldn't be paired." );
266265 }
267266
268267 // By default, batch is the 0 - th but chatglm uses 1 - st dimension as batch
0 commit comments