cleanup unused header in sgl_kernel (#2986)

sgl-kernel/src/sgl-kernel/csrc/cutlass_extensions/epilogue/epilogue_per_row_per_col_scale.h

@@ -3,11 +3,8 @@
 
 #pragma once
 
-#include "cutlass/arch/memory.h"
-#include "cutlass/arch/memory_sm75.h"
-#include "cutlass/cutlass.h"
-#include "cutlass/fast_math.h"
-#include "cutlass/numeric_conversion.h"
+#include <cutlass/arch/memory.h>
+#include <cutlass/numeric_conversion.h>
 
 namespace cutlass {
 namespace epilogue {

sgl-kernel/src/sgl-kernel/csrc/cutlass_extensions/gemm/gemm_universal_base_compat.h

@@ -2,16 +2,9 @@
 // https://github.com/NVIDIA/TensorRT-LLM/blob/be1788106245496872d18e702978e59b6bfd50e0/cpp/tensorrt_llm/cutlass_extensions/include/cutlass_extensions/gemm/device/gemm_universal_base_compat.h
 #pragma once
 
-#include "cutlass/arch/arch.h"
-#include "cutlass/cutlass.h"
-#include "cutlass/device_kernel.h"
-#include "cutlass/gemm/device/default_gemm_configuration.h"
-#include "cutlass/gemm/gemm.h"
-#include "cutlass/gemm/kernel/default_gemm_universal.h"
-#include "cutlass/gemm/kernel/gemm_universal.h"
-#include "cutlass/gemm/threadblock/threadblock_swizzle.h"
-#include "cutlass/numeric_types.h"
-#include "cutlass/trace.h"
+#include <cutlass/cutlass.h>
+#include <cutlass/device_kernel.h>
+#include <cutlass/trace.h>
 
 ////////////////////////////////////////////////////////////////////////////////
 

sgl-kernel/src/sgl-kernel/csrc/cutlass_extensions/gemm/gemm_with_epilogue_visitor.h

@@ -3,14 +3,11 @@
 
 #pragma once
 
-#include "cutlass/complex.h"
-#include "cutlass/cutlass.h"
-#include "cutlass/fast_math.h"
-#include "cutlass/gemm/gemm.h"
-#include "cutlass/matrix_coord.h"
-#include "cutlass/semaphore.h"
-#include "cutlass/trace.h"
-#include "cutlass_extensions/epilogue/epilogue_per_row_per_col_scale.h"
+#include <cutlass/complex.h>
+#include <cutlass/cutlass.h>
+#include <cutlass/fast_math.h>
+#include <cutlass/matrix_coord.h>
+#include <cutlass/trace.h>
 
 /////////////////////////////////////////////////////////////////////////////////////////////////
 

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

+#include <vector>
+
 #include "utils.hpp"
 
 // trt_reduce

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

@@ -3,9 +3,6 @@
 #include <c10/cuda/CUDAStream.h>
 
 #include <cassert>
-#include <iostream>
-#include <sstream>
-#include <unordered_map>
 
 #include "trt_reduce_internal.cuh"