support GQA

Author: Kotomi-Du

onnxruntime/core/providers/openvino/backends/basic_backend.h

@@ -57,6 +57,7 @@ struct OnnxToOvNetworkBindings {
       "beam_idx",
       "past_key_values",
       "present",
+      "total_seq_len",
   };
 
   OnnxToOvNetworkBindings(OVExeNetwork& exec_network, SubGraphContext& subgraph_context, SessionContext& session_context) {

onnxruntime/core/providers/openvino/ov_versions/data_ops.cc

@@ -96,6 +96,7 @@ std::vector<SupportedOp> supported_op_mode = {
     {"Atanh", V_2020_4, {"CPU"}},
     {"Atanh", V_2022_1, {"GPU"}},
     {"Attention", V_2023_0, {"CPU", "GPU"}},
+    {"GroupQueryAttention", V_2025_1, {"CPU", "GPU"}},
     {"AveragePool", V_2020_4, {"CPU", "GPU"}},
     {"BatchNormalization", V_2020_4, {"CPU", "GPU"}},
     {"BiasGelu", V_2023_0, {"CPU", "GPU"}},