|
||
---|---|---|
.. | ||
ResNet50 | ||
README.md |
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