use fast math for per_token_group_quant_8bit. (#9177)

Co-authored-by: Zhang Kaihong <zhangkaihong.zkh@alibaba-inc.com>

sgl-kernel/CMakeLists.txt

@@ -301,6 +301,8 @@ target_include_directories(common_ops PRIVATE
     ${repo-cutlass_SOURCE_DIR}/examples/common
     ${repo-flash-attention_SOURCE_DIR}/csrc/flash_attn/src
 )
+set_source_files_properties("csrc/gemm/per_token_group_quant_8bit" PROPERTIES COMPILE_OPTIONS "--use_fast_math")
+
 
 find_package(Python3 COMPONENTS Interpreter REQUIRED)
 execute_process(