fix example

example/32_batched_gemm_scale_softmax_gemm/grouped_multihead_attention_backward_fp16.cpp

@@ -287,15 +287,15 @@ int run(int argc, char* argv[])
     std::vector<DeviceGemmInstance::ProblemDesc> problem_descs;
 
     using DeviceMemPtr = std::unique_ptr<DeviceMem>;
-    std::vector<const DataType*> p_q;
-    std::vector<const DataType*> p_k;
-    std::vector<const DataType*> p_v;
-    std::vector<const DataType*> p_y;
-    std::vector<const LSEDataType*> p_lse;
-    std::vector<DataType*> p_qgrad;
-    std::vector<DataType*> p_kgrad;
-    std::vector<DataType*> p_vgrad;
-    std::vector<const DataType*> p_ygrad;
+    std::vector<const void*> p_q;
+    std::vector<const void*> p_k;
+    std::vector<const void*> p_v;
+    std::vector<const void*> p_y;
+    std::vector<const void*> p_lse;
+    std::vector<void*> p_qgrad;
+    std::vector<void*> p_kgrad;
+    std::vector<void*> p_vgrad;
+    std::vector<const void*> p_ygrad;
 
     std::vector<Tensor<DataType>> q_g_m_ks;
     std::vector<Tensor<DataType>> k_g_n_ks;
@@ -517,15 +517,15 @@ int run(int argc, char* argv[])
         kgrad_tensors_device.back()->SetZero();
         vgrad_tensors_device.back()->SetZero();
         ygrad_tensors_device.back()->ToDevice(ygrad_gs_ms_os.data());
-        p_q.push_back(static_cast<DataType*>(q_tensors_device.back()->GetDeviceBuffer()));        
-        p_k.push_back(static_cast<DataType*>(k_tensors_device.back()->GetDeviceBuffer()));        
-        p_v.push_back(static_cast<DataType*>(v_tensors_device.back()->GetDeviceBuffer()));        
-        p_y.push_back(static_cast<DataType*>(y_tensors_device.back()->GetDeviceBuffer()));        
-        p_lse.push_back(static_cast<LSEDataType*>(lse_tensors_device.back()->GetDeviceBuffer()));
-        p_kgrad.push_back(static_cast<DataType*>(kgrad_tensors_device.back()->GetDeviceBuffer()));
-        p_vgrad.push_back(static_cast<DataType*>(vgrad_tensors_device.back()->GetDeviceBuffer()));
-        p_ygrad.push_back(static_cast<DataType*>(ygrad_tensors_device.back()->GetDeviceBuffer()));
-        p_qgrad.push_back(static_cast<DataType*>(qgrad_tensors_device.back()->GetDeviceBuffer()));
+        p_q.push_back(q_tensors_device.back()->GetDeviceBuffer());        
+        p_k.push_back(k_tensors_device.back()->GetDeviceBuffer());        
+        p_v.push_back(v_tensors_device.back()->GetDeviceBuffer());        
+        p_y.push_back(y_tensors_device.back()->GetDeviceBuffer());        
+        p_lse.push_back(lse_tensors_device.back()->GetDeviceBuffer());
+        p_kgrad.push_back(kgrad_tensors_device.back()->GetDeviceBuffer());
+        p_vgrad.push_back(vgrad_tensors_device.back()->GetDeviceBuffer());
+        p_ygrad.push_back(ygrad_tensors_device.back()->GetDeviceBuffer());
+        p_qgrad.push_back(qgrad_tensors_device.back()->GetDeviceBuffer());
     }
     auto argument = gemm.MakeArgument(
         p_q,