-
Notifications
You must be signed in to change notification settings - Fork 91
/
Copy pathsample_training_runs.sh
72 lines (52 loc) · 7.26 KB
/
sample_training_runs.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
################################################################################################################
################################################ ResNet RUNS ###################################################
### ResNet50 on CUB200-2011: MarginLoss, TripletLoss + Semihard, ProxyNCA, N-Pair
python Standard_Training.py --gpu 0 --savename resnet_cub_margin_dist --dataset cub200 --n_epochs 50 --tau 30 --loss marginloss --sampling distance
python Standard_Training.py --gpu 0 --savename resnet_cub_triplet_semi --dataset cub200 --n_epochs 50 --tau 30 --loss triplet --sampling semihard
python Standard_Training.py --gpu 0 --savename resnet_cub_proxynca_none --dataset cub200 --n_epochs 40 --tau 25 --loss proxynca
python Standard_Training.py --gpu 0 --savename resnet_cub_npair_none --dataset cub200 --n_epochs 55 --tau 35 --loss npair
### ResNet50 on CARS196: MarginLoss, TripletLoss + Semihard, ProxyNCA, N-Pair
python Standard_Training.py --gpu 0 --savename resnet_cars_margin_dist --dataset cars196 --n_epochs 70 --tau 45 --loss marginloss --sampling distance
python Standard_Training.py --gpu 0 --savename resnet_cars_triplet_semi --dataset cars196 --n_epochs 70 --tau 45 --loss triplet --sampling semihard
python Standard_Training.py --gpu 0 --savename resnet_cars_proxynca_none --dataset cars196 --n_epochs 40 --tau 25 --loss proxynca
python Standard_Training.py --gpu 0 --savename resnet_cars_npair_none --dataset cars196 --n_epochs 85 --tau 55 --loss npair
### ResNet50 on Stanford Online-Products: MarginLoss, TripletLoss + Semihard, N-Pair
python Standard_Training.py --gpu 0 --savename resnet_op_margin_dist --dataset online_products --n_epochs 40 --tau 30 --loss marginloss --sampling distance
python Standard_Training.py --gpu 0 --savename resnet_op_triplet_semi --dataset online_products --n_epochs 40 --tau 30 --loss triplet --sampling semihard
python Standard_Training.py --gpu 0 --savename resnet_op_npair_none --dataset online_products --n_epochs 25 --tau 15 20 --loss npair
### ResNet50 on In-Shop Clothes: MarginLoss, TripletLoss + Semihard, N-Pair
python Standard_Training.py --gpu 0 --savename resnet_inshop_margin_dist --dataset in-shop --n_epochs 40 --tau 25 --loss marginloss --sampling distance
python Standard_Training.py --gpu 0 --savename resnet_inshop_triplet_semi --dataset in-shop --n_epochs 40 --tau 25 --loss triplet --sampling semihard
python Standard_Training.py --gpu 0 --savename resnet_inshop_npair_none --dataset in-shop --n_epochs 40 --tau 25 --loss npair
# ### (optional) ResNet50 on Vehicle-ID
# python Standard_Training.py --gpu 0 --savename resnet_vehicle_margin_dist --dataset vehicle_id --n_epochs 40 --tau 25 --loss marginloss --sampling distance
# python Standard_Training.py --gpu 0 --savename resnet_vehicle_triplet_semi --dataset vehicle_id --n_epochs 40 --tau 25 --loss triplet --sampling semihard
# python Standard_Training.py --gpu 0 --savename resnet_vehicle_npair_none --dataset vehicle_id --n_epochs 40 --tau 25 --loss npair
################################################################################################################
############################################# GoogLeNet RUNS ###################################################
### GoogLeNet on CUB200-2011: MarginLoss, TripletLoss + Semihard, ProxyNCA, N-Pair
python Standard_Training.py --gpu 0 --arch googlenet --lr 0.0001 --embed_dim 512 --savename gnet_cars_margin_dist --dataset cars196 --n_epochs 70 --tau 30 50 --loss marginloss --sampling distance
python Standard_Training.py --gpu 0 --arch googlenet --lr 0.0001 --embed_dim 512 --savename gnet_cars_triplet_semi --dataset cars196 --n_epochs 70 --tau 30 50 --loss triplet --sampling semihard
python Standard_Training.py --gpu 0 --arch googlenet --lr 0.0001 --embed_dim 512 --savename gnet_cars_proxynca_none --dataset cars196 --n_epochs 40 --tau 20 30 --loss proxynca
python Standard_Training.py --gpu 0 --arch googlenet --lr 0.0001 --embed_dim 512 --savename gnet_cars_npair_none --dataset cars196 --n_epochs 85 --tau 40 65 --loss npair
### GoogLeNet on CARS196: MarginLoss, TripletLoss + Semihard, ProxyNCA, N-Pair
python Standard_Training.py --gpu 0 --arch googlenet --lr 0.0001 --embed_dim 512 --savename gnet_cub_margin_dist --dataset cub200 --n_epochs 60 --tau 10 20 30 40 50 --loss marginloss --sampling distance
python Standard_Training.py --gpu 0 --arch googlenet --lr 0.0001 --embed_dim 512 --savename gnet_cub_triplet_semi --dataset cub200 --n_epochs 70 --tau 15 40 60 --loss triplet --sampling semihard
python Standard_Training.py --gpu 0 --arch googlenet --lr 0.0001 --embed_dim 512 --savename gnet_cub_proxynca_none --dataset cub200 --n_epochs 25 --tau 7 17 --loss proxynca
python Standard_Training.py --gpu 0 --arch googlenet --lr 0.0001 --embed_dim 512 --savename gnet_cub_npair_none --dataset cub200 --n_epochs 150 --tau 100 --loss npair
### GoogLeNet on Online Products: MarginLoss, TripletLoss + Semihard, N-Pair
python Standard_Training.py --gpu 0 --arch googlenet --lr 0.0001 --embed_dim 512 --savename gnet_op_margin_dist --dataset online_products --n_epochs 20 --tau 2 6 --loss marginloss --sampling distance
python Standard_Training.py --gpu 0 --arch googlenet --lr 0.0001 --embed_dim 512 --savename gnet_op_triplet_semi --dataset online_products --n_epochs 40 --tau 10 --loss triplet --sampling semihard
python Standard_Training.py --gpu 0 --arch googlenet --lr 0.0001 --embed_dim 512 --savename gnet_op_npair_none --dataset online_products --n_epochs 20 --tau 2 6 --loss npair
### GoogLeNet on In-Shop Clothes: MarginLoss, TripletLoss + Semihard, N-Pair
python Standard_Training.py --gpu 0 --arch googlenet --lr 0.0001 --embed_dim 512 --savename gnet_inshop_margin_dist --dataset in-shop --n_epochs 40 --tau 25 --loss marginloss --sampling distance
python Standard_Training.py --gpu 0 --arch googlenet --lr 0.0001 --embed_dim 512 --savename gnet_inshop_triplet_semi --dataset in-shop --n_epochs 40 --tau 25 --loss triplet --sampling semihard
python Standard_Training.py --gpu 0 --arch googlenet --lr 0.0001 --embed_dim 512 --savename gnet_inshop_npair_none --dataset in-shop --n_epochs 40 --tau 25 --loss npair
# ### (optional) GoogLeNet on Vehicle-ID
# python Standard_Training.py --gpu 0 --arch googlenet --lr 0.0001 --embed_dim 512 --savename gnet_vehicle_margin_dist --dataset vehicle_id --n_epochs 40 --tau 25 --loss marginloss --sampling distance
# python Standard_Training.py --gpu 0 --arch googlenet --lr 0.0001 --embed_dim 512 --savename gnet_vehicle_triplet_semi --dataset vehicle_id --n_epochs 40 --tau 25 --loss triplet --sampling semihard
# python Standard_Training.py --gpu 0 --arch googlenet --lr 0.0001 --embed_dim 512 --savename gnet_vehicle_npair_none --dataset vehicle_id --n_epochs 40 --tau 25 --loss npair
########################################################################################
#################################### Special RUNS ######################################
python Standard_Training.py --gpu 0 --savename resnet_cars_margin_dist --dataset cars196 --n_epochs 70 --tau 45 --loss marginloss --sampling distance --distance_measure
python Standard_Training.py --gpu 0 --savename resnet_cars_margin_dist --dataset cars196 --n_epochs 70 --tau 45 --loss marginloss --sampling distance --grad_measure