PaddleClas/test_tipc/static
liuyuang fd2d4faac4 more data for bs256 2022-06-07 12:32:03 +08:00
..
ResNet50 more data for bs256 2022-06-07 12:32:03 +08:00
README.md add static graph ResNet50 for benchmark 2022-02-21 08:37:02 +00:00

README.md

PaddleClas 下静态图benchmark模型执行说明

静态图benchmark测试脚本说明

目录说明

Docker 运行环境

docker image: registry.baidubce.com/paddlepaddle/paddle:latest-dev-cuda11.2-cudnn8-gcc82 paddle = 2.2.2 python = 3.7

运行benchmark测试步骤

git clone https://github.com/PaddlePaddle/PaddleClas.git
cd PaddleClas

准备数据

bash test_tipc/static/${model_item}/benchmark_common/prepare.sh

运行模型

单卡自动运行打开Profiling

export CUDA_VISIBLE_DEVICES=0
bash  test_tipc/static/${model_item}/N1C1/${shell_name}.sh

多卡

export CUDA_VISIBLE_DEVICES=0,1,2,3,4,5,6,7
bash test_tipc/static/${model_item}/N1C8/${shell_name}.sh