After testing ，I found pool performance on the fpga-card，is that normal？

[mayflyfy]

at first I test yolov2-608, and found that U200 performance is lower than NVIDIA 2080Ti

and then I change the input image size for yolov2,
then made the form，

U200 inference speed will only better when the input image size is lower than 224，

is that normal？

thanks！

[mayflyfy]

输入层尺寸 means input image size 推理用时 means inference time(ms/img)

[zyxcambridge]

at first I test yolov2-608, and found that U200 performance is lower than NVIDIA 2080Ti

and then I change the input image size for yolov2, then made the form，

U200 inference speed will only better when the input image size is lower than 224，

is that normal？

thanks！

你这边没有用docker 跑起来了，是吧

[mayflyfy]

at first I test yolov2-608, and found that U200 performance is lower than NVIDIA 2080Ti and then I change the input image size for yolov2, then made the form， U200 inference speed will only better when the input image size is lower than 224， is that normal？ thanks！

你这边没有用docker 跑起来了，是吧

我用的是xilinx官方的docker，镜像名称是xilinx-ml-suite-ubuntu-16.04-xrt-2018.2-caffe-1.0-mls-1.5 请问你测试的推理速度如何？

[zyxcambridge]

at first I test yolov2-608, and found that U200 performance is lower than NVIDIA 2080Ti and then I change the input image size for yolov2, then made the form， U200 inference speed will only better when the input image size is lower than 224， is that normal？ thanks！

你这边没有用docker 跑起来了，是吧

我用的是xilinx官方的docker，镜像名称是xilinx-ml-suite-ubuntu-16.04-xrt-2018.2-caffe-1.0-mls-1.5 请问你测试的推理速度如何？

我也用的是xilink的官方docker xilinx-ml-suite-ubuntu-16.04-xrt-2018.2-caffe-1.0-mls-1.5 在 aws f1 上 ，但是各种环境不兼容；

请教一下： 你这边是本地机子上 装了一块 u200的 FPGA卡，才跑起来的是吧；

另外我这边 想知道一下 yolov2 和yolov3的MAP ，感谢， 顺便来个微信 吧，zhangyixin395104 我的微信

[llt19903767731]

at first I test yolov2-608, and found that U200 performance is lower than NVIDIA 2080Ti and then I change the input image size for yolov2, then made the form， U200 inference speed will only better when the input image size is lower than 224， is that normal？ thanks！

你这边没有用docker 跑起来了，是吧

我用的是xilinx官方的docker，镜像名称是xilinx-ml-suite-ubuntu-16.04-xrt-2018.2-caffe-1.0-mls-1.5 请问你测试的推理速度如何？

这个镜像是XRT 2018.2版本的吧，如果本机装Xilinx的XRT是2019年的，好像没有对应的镜像吧？我在仓库好像没找到。