Skip to content
体验新版
项目
组织
正在加载...
登录
切换导航
打开侧边栏
PaddlePaddle
PaddleDetection
提交
6bf1b443
P
PaddleDetection
项目概览
PaddlePaddle
/
PaddleDetection
大约 1 年 前同步成功
通知
695
Star
11112
Fork
2696
代码
文件
提交
分支
Tags
贡献者
分支图
Diff
Issue
184
列表
看板
标记
里程碑
合并请求
40
Wiki
0
Wiki
分析
仓库
DevOps
项目成员
Pages
P
PaddleDetection
项目概览
项目概览
详情
发布
仓库
仓库
文件
提交
分支
标签
贡献者
分支图
比较
Issue
184
Issue
184
列表
看板
标记
里程碑
合并请求
40
合并请求
40
Pages
分析
分析
仓库分析
DevOps
Wiki
0
Wiki
成员
成员
收起侧边栏
关闭侧边栏
动态
分支图
创建新Issue
提交
Issue看板
未验证
提交
6bf1b443
编写于
10月 15, 2021
作者:
F
Feng Ni
提交者:
GitHub
10月 15, 2021
浏览文件
操作
浏览文件
下载
电子邮件补丁
差异文件
[benchmark] add jde/fairmot model training benchmark (#4307)
上级
e14387f8
变更
4
隐藏空白更改
内联
并排
Showing
4 changed file
with
20 addition
and
9 deletion
+20
-9
benchmark/README.md
benchmark/README.md
+2
-2
benchmark/prepare.sh
benchmark/prepare.sh
+7
-0
benchmark/run_all.sh
benchmark/run_all.sh
+4
-2
benchmark/run_benchmark.sh
benchmark/run_benchmark.sh
+7
-5
未找到文件。
benchmark/README.md
浏览文件 @
6bf1b443
...
...
@@ -33,7 +33,7 @@ bash benchmark/run_all.sh
### 运行指定模型
*
Usage:bash run_benchmark.sh ${run_mode} ${batch_size} ${fp_item} ${max_epoch} ${model_name}
*
model_name: faster_rcnn, fcos, deformable_detr, gfl
*
model_name: faster_rcnn, fcos, deformable_detr, gfl
, hrnet, higherhrnet, solov2, jde, fairmot
```
git clone https://github.com/PaddlePaddle/PaddleDetection.git
cd PaddleDetection
...
...
@@ -42,5 +42,5 @@ bash benchmark/prepare.sh
# 单卡
CUDA_VISIBLE_DEVICES=0 bash benchmark/run_benchmark.sh sp 2 fp32 1 faster_rcnn
# 多卡
CUDA_VISIBLE_DEVICES=0,1,2,3,4,5,6,7 bash run_benchmark.sh mp 2 fp32 1 faster_rcnn
CUDA_VISIBLE_DEVICES=0,1,2,3,4,5,6,7 bash
benchmark/
run_benchmark.sh mp 2 fp32 1 faster_rcnn
```
benchmark/prepare.sh
浏览文件 @
6bf1b443
...
...
@@ -2,8 +2,15 @@
pip3.7
install
-U
pip Cython
pip3.7
install
-r
requirements.txt
mv
./dataset/coco/download_coco.py
.
&&
rm
-rf
./dataset/coco/
*
&&
mv
./download_coco.py ./dataset/coco/
# prepare lite train data
wget
-nc
-P
./dataset/coco/ https://paddledet.bj.bcebos.com/data/coco_benchmark.tar
cd
./dataset/coco/
&&
tar
-xvf
coco_benchmark.tar
&&
mv
-u
coco_benchmark/
*
.
rm
-rf
coco_benchmark/
rm
-rf
./dataset/mot/
*
# prepare mot mini train data
wget
-nc
-P
./dataset/mot/ https://paddledet.bj.bcebos.com/data/mot_benchmark.tar
cd
./dataset/mot/
&&
tar
-xvf
mot_benchmark.tar
&&
mv
-u
mot_benchmark/
*
.
rm
-rf
mot_benchmark/
benchmark/run_all.sh
浏览文件 @
6bf1b443
...
...
@@ -8,7 +8,7 @@
# run prepare.sh
bash benchmark/prepare.sh
model_name_list
=(
faster_rcnn fcos deformable_detr gfl
)
model_name_list
=(
faster_rcnn fcos deformable_detr gfl
hrnet higherhrnet solov2 jde fairmot
)
fp_item_list
=(
fp32
)
max_epoch
=
1
...
...
@@ -22,6 +22,8 @@ for model_name in ${model_name_list[@]}; do
hrnet
)
bs_list
=(
64 160
)
;;
higherhrnet
)
bs_list
=(
20 24
)
;;
solov2
)
bs_list
=(
2 4
)
;;
jde
)
bs_list
=(
4 14
)
;;
fairmot
)
bs_list
=(
6 22
)
;;
*
)
echo
"wrong model_name"
;
exit
1
;
esac
for
bs_item
in
${
bs_list
[@]
}
...
...
@@ -29,7 +31,7 @@ for model_name in ${model_name_list[@]}; do
echo
"index is speed, 1gpus, begin,
${
model_name
}
"
run_mode
=
sp
CUDA_VISIBLE_DEVICES
=
0 bash benchmark/run_benchmark.sh
${
run_mode
}
${
bs_item
}
\
${
fp_item
}
${
max_epoch
}
${
model_name
}
# (5min)
${
fp_item
}
${
max_epoch
}
${
model_name
}
sleep
60
echo
"index is speed, 8gpus, run_mode is multi_process, begin,
${
model_name
}
"
...
...
benchmark/run_benchmark.sh
浏览文件 @
6bf1b443
#!/usr/bin/env bash
set
-xe
# Usage:CUDA_VISIBLE_DEVICES=0 bash run_benchmark.sh ${run_mode} ${batch_size} ${fp_item} ${max_epoch} ${model_name}
# Usage:CUDA_VISIBLE_DEVICES=0 bash
benchmark/
run_benchmark.sh ${run_mode} ${batch_size} ${fp_item} ${max_epoch} ${model_name}
python
=
"python3.7"
# Parameter description
function
_set_params
(){
run_mode
=
${
1
:-
"sp"
}
# sp|mp
batch_size
=
${
2
:-
"2"
}
#
batch_size
=
${
2
:-
"2"
}
fp_item
=
${
3
:-
"fp32"
}
# fp32|fp16
max_epoch
=
${
4
:-
"1"
}
#
max_epoch
=
${
4
:-
"1"
}
model_name
=
${
5
:-
"model_name"
}
run_log_path
=
${
TRAIN_LOG_DIR
:-
$(
pwd
)
}
# TRAIN_LOG_DIR
run_log_path
=
${
TRAIN_LOG_DIR
:-
$(
pwd
)
}
device
=
${
CUDA_VISIBLE_DEVICES
//,/
}
arr
=(
${
device
}
)
...
...
@@ -29,6 +29,8 @@ function _train(){
hrnet
)
model_yml
=
"configs/keypoint/hrnet/hrnet_w32_256x192.yml"
;;
higherhrnet
)
model_yml
=
"configs/keypoint/higherhrnet/higherhrnet_hrnet_w32_512.yml"
;;
solov2
)
model_yml
=
"configs/solov2/solov2_r50_fpn_1x_coco.yml"
;;
jde
)
model_yml
=
"configs/mot/jde/jde_darknet53_30e_1088x608.yml"
;;
fairmot
)
model_yml
=
"configs/mot/fairmot/fairmot_dla34_30e_1088x608.yml"
;;
*
)
echo
"Undefined model_name"
;
exit
1
;
esac
...
...
@@ -50,7 +52,7 @@ function _train(){
log_parse_file
=
"mylog/workerlog.0"
;;
*
)
echo
"choose run_mode(sp or mp)"
;
exit
1
;
esac
#
timeout
15m
${
train_cmd
}
>
${
log_file
}
2>&1
if
[
$?
-ne
0
]
;
then
echo
-e
"
${
train_cmd
}
, FAIL"
...
...
编辑
预览
Markdown
is supported
0%
请重试
或
添加新附件
.
添加附件
取消
You are about to add
0
people
to the discussion. Proceed with caution.
先完成此消息的编辑!
取消
想要评论请
注册
或
登录