-
Notifications
You must be signed in to change notification settings - Fork 4
/
avs_task.sh
260 lines (220 loc) · 9.46 KB
/
avs_task.sh
1
2
3
4
5
6
7
8
9
10
11
12
13
14
15
16
17
18
19
20
21
22
23
24
25
26
27
28
29
30
31
32
33
34
35
36
37
38
39
40
41
42
43
44
45
46
47
48
49
50
51
52
53
54
55
56
57
58
59
60
61
62
63
64
65
66
67
68
69
70
71
72
73
74
75
76
77
78
79
80
81
82
83
84
85
86
87
88
89
90
91
92
93
94
95
96
97
98
99
100
101
102
103
104
105
106
107
108
109
110
111
112
113
114
115
116
117
118
119
120
121
122
123
124
125
126
127
128
129
130
131
132
133
134
135
136
137
138
139
140
141
142
143
144
145
146
147
148
149
150
151
152
153
154
155
156
157
158
159
160
161
162
163
164
165
166
167
168
169
170
171
172
173
174
175
176
177
178
179
180
181
182
183
184
185
186
187
188
189
190
191
192
193
194
195
196
197
198
199
200
201
202
203
204
205
206
207
208
209
210
211
212
213
214
215
216
217
218
219
220
221
222
223
224
225
226
227
228
229
230
231
232
233
234
235
236
237
238
239
240
241
242
243
244
245
246
247
248
249
250
251
252
253
254
255
256
257
258
259
260
#!/bin/sh
<<COMMENT
这个脚本是把训练，生成测试文件，生成 avs检索 结果三部分结合起来，自动进行。
注意：
1. 更改模型在 config 文件中更改。
2. random_seed 是传入 do_train.py 文件的一个参数，如果没有使用可以修改或者删除。
3. result_file 是存储结果的文件名。
3. parm_adjust_configs 是传入 config 文件的一个参数，如果没有使用可以修改或者删除。
COMMENT
# *************************萌萌哒******************************
# 并行参数
Nproc=1 # 可同时运行的最大作业数
devices=()
# shellcheck disable=SC2113
function PushQue { # 将PID压入队列
Que="$Que $1"
Nrun=$(($Nrun+1))
}
function GenQue { # 更新队列
OldQue=$Que
Que=""; Nrun=0
for PID in $OldQue; do
if [[ -d /proc/$PID ]]; then
PushQue $PID
fi
done
}
function ChkQue { # 检查队列
OldQue=$Que
for PID in $OldQue; do
if [[ ! -d /proc/$PID ]] ; then
GenQue; break
fi
done
}
function paralle {
PID=$!
PushQue $PID
while [[ $Nrun -ge ${Nproc} ]]; do
ChkQue
sleep 1
done
}
# *************************萌萌哒******************************
path_shell=`pwd`
cd ..
path_w2vvpp=`pwd`
rootpath="$HOME/hf_code/VisualSearch"
trainCollection=""
valCollection=""
val_set='' # setA
testCollection=''
txt_feature_task2='no'
config=''
batch_size=128
workers=2
overwrite=0
pretrained_file_path='None' # 默认没有 pretrained_file
only_train=0
save_mean_last=0
trainCollection2='None'
random_seeds=(2) # 初始化随机数种子
#parm_adjust_configs=()
model_prefix_="runs_"
result_file="$path_w2vvpp/result_log/result_${model_prefix_}_${config}.txt"
# 读取输入的参数
GETOPT_ARGS=$(getopt -o l: -al rootpath:,trainCollection:,valCollection:,val_set:,\
testCollection:,config:,batch_size:,overwrite:,devices:,Nproc:,random_seeds:,\
pretrained_file_path:,parm_adjust_configs:,\
model_prefix_:,result_file:,save_mean_last:,workers:,trainCollection2:,only_train: -- "$@") # , 后一定不要有空格
eval set -- "$GETOPT_ARGS"
#获取参数
while [ -n "$1" ]
do
case "$1" in
--rootpath) rootpath=$2; shift 2;;
--trainCollection) trainCollection=$2; shift 2;;
--trainCollection2) trainCollection2=$2; shift 2;;
--valCollection) valCollection=$2; shift 2;;
--val_set) val_set=$2; shift 2;;
--testCollection) testCollection=$2; shift 2;;
--config) config=$2; shift 2;;
--batch_size) batch_size=$(($2)); shift 2;;
--overwrite) overwrite=$2; shift 2;;
--devices) devices_temp=$2; shift 2;;
--Nproc) Nproc=$(($2)); shift 2;;
--parm_adjust_configs) parm_adjust_configs=$2; shift 2;;
--random_seeds) random_seeds=$2; shift 2;;
--pretrained_file_path) pretrained_file_path=$2; shift 2;;
--model_prefix_) model_prefix_=$2; shift 2;;
--result_file) result_file=$2; shift 2;;
--only_train) only_train=$(($2)); shift 2;;
--save_mean_last) save_mean_last=$(($2)); shift 2;;
--workers) workers=$(($2)); shift 2;;
--) break ;;
*) echo $1,$2; break ;;
esac
done
echo "result_file：$result_file, devices: ${devices[*]} , config: $config , parm_adjust_configs: ${parm_adjust_configs},
Nproc: ${Nproc} "
for each in ${devices_temp[*]}
do
devices[${#devices[@]}]=$each
done
#exit 0
# ****************************************
# 训练
# shellcheck disable=SC2039
device_index=0
for random_seed in ${random_seeds[*]}
do
for parm_adjust_config in ${parm_adjust_configs[*]}
do
model_prefix="${model_prefix_}${parm_adjust_config}_seed_${random_seed}"
if [[ ${pretrained_file_path} != 'None' ]]; then
pretrained_file_path_model="${pretrained_file_path}/${model_prefix}/model_best.pth.tar"
else
pretrained_file_path_model='None'
fi
# echo "let us train ${trainCollection2}"
device=${devices[device_index]}
let device_index="($device_index + 1) % ${#devices[*]}"
if [[ $Nproc -gt 1 ]]; then
echo "$trainCollection $valCollection --rootpath $rootpath --config $config --val_set $val_set --model_prefix $model_prefix"
python do_trainer.py $trainCollection $valCollection \
--rootpath $rootpath --config_name $config --val_set $val_set --model_prefix $model_prefix \
--batch_size $batch_size --workers $workers --device $device --overwrite $overwrite \
--parm_adjust_config $parm_adjust_config --pretrained_file_path ${pretrained_file_path_model} \
--random_seed $random_seed --save_mean_last ${save_mean_last} --trainCollection2 ${trainCollection2} \
& paralle
else
python do_trainer.py $trainCollection $valCollection \
--rootpath $rootpath --config_name $config --val_set $val_set --model_prefix $model_prefix \
--batch_size $batch_size --workers $workers --device $device --overwrite $overwrite \
--parm_adjust_config $parm_adjust_config --pretrained_file_path ${pretrained_file_path_model} \
--random_seed $random_seed --save_mean_last ${save_mean_last} --trainCollection2 ${trainCollection2}
fi
done
done
if [[ ${only_train} = 1 ]]
then
exit 0
fi
wait
## ******************萌萌哒分割线***********************
# 测试与输出结果
overwrite=1
echo "overwrite" ${overwrite}
batch_size=1024
# create result_file directory
result_file_path=${path_w2vvpp}/${result_file}
mkdir -p ${result_file_path%/*} # 创建文件目录以及父目录
model_best_names=('model_best.pth.tar' 'mean_last10.pth.tar')
for model_best_name in ${model_best_names[*]}
do
for random_seed in ${random_seeds[*]}
do
for parm_adjust_config in ${parm_adjust_configs[*]}
do
# overwrite=0
model_prefix="${model_prefix_}${parm_adjust_config}_seed_${random_seed}"
if [[ $val_set = 'no' ]]
then
model_path=$rootpath/$trainCollection/w2vvpp_train/$valCollection/$config/$model_prefix/${model_best_name}
sim_name=$trainCollection/$valCollection/$config/$model_prefix/${model_best_name}
else
model_path=$rootpath/$trainCollection/w2vvpp_train/$valCollection/$val_set/$config/$model_prefix/${model_best_name}
sim_name=$trainCollection/$valCollection/$val_set/$config/$model_prefix/${model_best_name}
fi
# tv16,tv17,tv18
query_sets=tv16.avs.txt,tv17.avs.txt,tv18.avs.txt
testCollection=iacc.3
python do_predictor.py $testCollection $model_path $sim_name \
--query_sets $query_sets \
--rootpath $rootpath --overwrite $overwrite --device $device \
--batch_size $batch_size --predict_result_file ${result_file_path} --num_workers $workers
# tv2019 2020 2021
testCollection=v3c1
query_sets=tv19.avs.txt,tv20.avs.txt,tv21.avs.txt,trecvid.progress.avs.txt
python do_predictor.py $testCollection $model_path $sim_name \
--query_sets $query_sets \
--rootpath $rootpath --overwrite $overwrite --device $device \
--batch_size $batch_size --predict_result_file ${result_file_path} --num_workers $workers
## ******************萌萌哒分割线***********************
# 评估表现
overwrite=1
cd tv_avs_eval
# shellcheck disable=SC2046
echo -e "\n" $(date "+%Y-%m-%d %H:%M:%S") "\t\c" | tee -a "${result_file_path}" # 输出到文件
echo -e ${model_path} ${parm_adjust_config//_/"\t"} "\t\c" | tee -a "${result_file_path}"
# tv16, 17, 18
for topic_set in "tv16" 'tv17' 'tv18'
do
test_collection="iacc.3"
score_file=$rootpath/$test_collection/SimilarityIndex/$topic_set.avs.txt/$sim_name/id.sent.score.txt
echo $score_file
python txt2xml.py $test_collection $score_file --edition $topic_set --priority 1 \
--etime 1.0 --desc "This run uses the top secret x-component" --rootpath $rootpath --overwrite $overwrite
python trec_eval.py ${score_file}.xml --rootpath $rootpath --collection $test_collection \
--edition $topic_set \
--overwrite $overwrite | tee -a "${result_file_path}"
# rm -rf $rootpath/$test_collection/SimilarityIndex/$topic_set.avs.txt/$sim_name/id.sent.score.txt
done
# tv19
for topic_set in 'tv19' 'tv20' 'tv21' 'trecvid.progress'
do
test_collection="v3c1"
score_file=$rootpath/$test_collection/SimilarityIndex/$topic_set.avs.txt/$sim_name/id.sent.score.txt
echo $score_file
python txt2xml.py $test_collection $score_file --edition $topic_set --priority 1 \
--etime 1.0 --desc "This run uses the top secret x-component" --rootpath $rootpath --overwrite $overwrite
python trec_eval.py ${score_file}.xml --rootpath $rootpath --collection $test_collection \
--edition $topic_set \
--overwrite $overwrite | tee -a "${result_file_path}"
# rm -rf $rootpath/$test_collection/SimilarityIndex/$topic_set.avs.txt/$sim_name/id.sent.score.txt
done
cd ..
done
done
done