OP Benchmark 测试算子性能的时候无法使用nsight system

[bapijun]

[PaddlePaddle OP Benchmark](https://github.com/PaddlePaddle/benchmark/tree/master/api)
在测试op的性能和精度的时候,无法正常的启动Nsight system,本人的显卡是3060,环境是docker环境下的latest-dev-cuda12.0-cudnn8.9-trt8.6-gcc12.2

[bapijun]

提示的错误是:
Running Error:
======== Warning: nvprof is not supported on devices with compute capability 8.0 and higher.
Use NVIDIA Nsight Systems for GPU tracing and CPU sampling and NVIDIA Nsight Compute for GPU profiling.
Refer https://developer.nvidia.com/tools-overview for more details.

怀疑是检查显卡的时候认定为低版本的显卡

[Xreki]

benchmark/api/common/launch.py

Lines 25 to 35 in bd98227

	def is_ampere_gpu():
	stdout, exit_code = system.run_command("nvidia-smi -L")
	if exit_code == 0:
	gpu_list = stdout.split("\n")
	if len(gpu_list) >= 1:
	#print(gpu_list[0])
	# GPU 0: NVIDIA A100-SXM4-40GB (UUID: xxxx)
	return gpu_list[0].find("A100") > 0
	return False

如上述代码所示，op benchmark已经支持了使用nsys得到GPU时间，但是当前代码里面判断是否ampere架构实现比较native，可通过简单的修改在3060上使用。