[Cherry-Pick][Op][Optimization]Kernel fusion: cast+sigmoid+bias+noauxtc(#7777) (#7832)

* [Op][Optimization]Kernel fusion: cast+sigmoid+bias+noauxtc (#7777)
[Cherry-Pick]

* Kernel fusion for blackwell and deepgemm backend in non-EPLB scenarios

* fix hard code in ep.py

Author: ShaneGZhu

custom_ops/gpu_ops/cpp_extensions.cc

@@ -693,6 +693,15 @@ std::vector<paddle::Tensor> NoauxTc(paddle::Tensor& scores,
                                     bool renormalize,
                                     float routed_scaling_factor);
 
+std::vector<paddle::Tensor> grouped_topk(
+    paddle::Tensor& gating_output,
+    paddle::Tensor& e_score_correction_bias,
+    int n_group,
+    int topk_group,
+    int topk,
+    bool renormalize,
+    float routed_scaling_factor);
+
 std::vector<paddle::Tensor> NoauxTcRedundant(
     paddle::Tensor& scores,
     paddle::Tensor& scores_with_bias,
@@ -1696,6 +1705,8 @@ PYBIND11_MODULE(fastdeploy_ops, m) {
 
   m.def("noaux_tc", &NoauxTc, "noaux_tc for Deepseekv3 MoE compute");
 
+  m.def("grouped_topk", &grouped_topk, "fused grouped topk for MoE routing");
+
   m.def("noaux_tc_redundant",
         &NoauxTcRedundant,
         "noaux_tc_redundant for MoE compute");