minor: update header and use pytest (#3054)

sgl-kernel/src/sgl-kernel/csrc/int8_gemm_kernel.cu

@@ -16,7 +16,7 @@
 #include "cutlass_extensions/epilogue/epilogue_per_row_per_col_scale.h"
 #include "cutlass_extensions/gemm/gemm_universal_base_compat.h"
 #include "cutlass_extensions/gemm/gemm_with_epilogue_visitor.h"
-#include "utils.hpp"
+#include "utils.h"
 
 using namespace cute;
 

sgl-kernel/src/sgl-kernel/csrc/moe_align_kernel.cu

@@ -6,7 +6,7 @@
 
 #include <THC/THCAtomics.cuh>
 
-#include "utils.hpp"
+#include "utils.h"
 
 #ifdef USE_ROCM
 #include <hip/hip_runtime.h>

sgl-kernel/src/sgl-kernel/csrc/sampling_scaling_penalties.cu

@@ -4,7 +4,7 @@
 
 #include <THC/THCAtomics.cuh>
 
-#include "utils.hpp"
+#include "utils.h"
 #include "vectorization.cuh"
 
 template <typename scalar_t>

sgl-kernel/src/sgl-kernel/csrc/sgl_kernel_ops.cu

@@ -1,6 +1,6 @@
 #include <vector>
 
-#include "utils.hpp"
+#include "utils.h"
 
 // trt_reduce
 using fptr_t = int64_t;

sgl-kernel/src/sgl-kernel/csrc/trt_reduce_internal.cuh

@@ -21,7 +21,7 @@
 #include <stdint.h>
 #include <torch/all.h>
 
-#include "utils.hpp"
+#include "utils.h"
 
 namespace trt_llm {
 constexpr size_t WARP_SIZE = 32;