[Fix]Fix index oob in get_group_gemm_starts kernel. (#8564)

sgl-kernel/csrc/moe/cutlass_moe_helper.cu

@@ -31,7 +31,7 @@ __global__ void get_group_gemm_starts(
     int* problem_sizes,
     int* problem_sizes_transpose,
     bool transpose = false) {
-  int expert_id = threadIdx.x;
+  int64_t expert_id = static_cast<int64_t>(threadIdx.x);
 
   if (expert_id >= gridDim.x * blockDim.x) {
     return;
@@ -46,11 +46,11 @@ __global__ void get_group_gemm_starts(
     problem_sizes_transpose[expert_id * 3 + 2] = k;
   }
 
-  int32_t expert_offset = expert_offsets[expert_id];
-  int a_stride = 0;
-  int b_stride = 0;
-  int a_scale_stride = 0;
-  int b_scale_stride = 0;
+  int64_t expert_offset = static_cast<int64_t>(expert_offsets[expert_id]);
+  int64_t a_stride = 0;
+  int64_t b_stride = 0;
+  int64_t a_scale_stride = 0;
+  int64_t b_scale_stride = 0;
   if (!transpose) {
     a_stride = expert_offset * k;
     b_stride = expert_id * k * n;