-
Notifications
You must be signed in to change notification settings - Fork 2
/
runall.sh
69 lines (56 loc) · 5.88 KB
/
runall.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
######################
# SimpleMLP #
######################
# splitMNIST, ensembling
python experiments/train_ex_model.py --model simple_mlp --scenario smnist --strategy model_average
python experiments/train_ex_model.py --model simple_mlp --scenario smnist --strategy model_ensemble
python experiments/train_ex_model.py --model simple_mlp --scenario smnist --strategy entropy_ensemble
# permutedMNIST, ensembling
python experiments/train_ex_model.py --model simple_mlp --scenario pmnist --strategy model_average
python experiments/train_ex_model.py --model simple_mlp --scenario pmnist --strategy model_ensemble
python experiments/train_ex_model.py --model simple_mlp --scenario pmnist --strategy entropy_ensemble
######################
# LeNet #
######################
# SplitMNIST, ensembling
python experiments/train_ex_model.py --model lenet --scenario smnist --strategy model_average
python experiments/train_ex_model.py --model lenet --scenario smnist --strategy model_ensemble
python experiments/train_ex_model.py --model lenet --scenario smnist --strategy entropy_ensemble
# SplitMNIST - aux FMNIST
python experiments/train_ex_model.py --model lenet --scenario smnist --strategy aux_data --lr 0.1 --epochs 10
# SplitMNIST, sampling-based naive
python experiments/train_ex_model.py --model lenet --scenario smnist --strategy buffer_naive --buffer_size 500 --lr 0.1 --epochs 1000
python experiments/train_ex_model.py --model lenet --scenario smnist --strategy minversion_naive --buffer_size 5000 --buffer_iter 10000 --buffer_tau 20 --buffer_wd 0.01 --lr 0.1
python experiments/train_ex_model.py --model lenet --scenario smnist --strategy dimpression_naive --buffer_size 5000 --buffer_iter 10000 --buffer_tau 20 --buffer_wd 0.01 --lr 0.1
# SplitMNIST, sampling-based cumulative
python experiments/train_ex_model.py --model lenet --scenario smnist --strategy buffer_cumulative --buffer_size 500 --lr 0.1 --epochs 1000
python experiments/train_ex_model.py --model lenet --scenario smnist --strategy minversion_cumulative --buffer_size 5000 --buffer_iter 10000 --buffer_tau 20 --buffer_wd 0.01 --lr 0.1
python experiments/train_ex_model.py --model lenet --scenario smnist --strategy dimpression_cumulative --buffer_size 5000 --buffer_iter 10000 --buffer_tau 20 --buffer_wd 0.01 --lr 0.1
# SplitMNIST, sampling-based replay
python experiments/train_ex_model.py --model lenet --scenario smnist --strategy buffer_replay --buffer_size 500 --lr 0.1 --mem_size 5000 --epochs 1000
python experiments/train_ex_model.py --model lenet --scenario smnist --strategy minversion_replay --buffer_size 5000 --buffer_iter 10000 --buffer_tau 20 --buffer_wd 0.01 --lr 0.1 --mem_size 50000
python experiments/train_ex_model.py --model lenet --scenario smnist --strategy dimpression_replay --buffer_size 5000 --buffer_iter 10000 --buffer_tau 20 --buffer_wd 0.01 --lr 0.1 --mem_size 50000
# JointMNIST, sampling-based naive
python experiments/train_ex_model.py --model lenet --scenario joint_mnist --strategy buffer_naive --buffer_size 500 --lr 0.1 --epochs 10000
python experiments/train_ex_model.py --model lenet --scenario joint_mnist --strategy dimpression_naive --buffer_size 50000 --buffer_iter 10000 --buffer_tau 20 --buffer_wd 0.01 --lr 0.1
python experiments/train_ex_model.py --model lenet --scenario joint_mnist --strategy minversion_naive --buffer_size 50000 --buffer_tau 20 --buffer_wd 0.01 --lr 0.1 --buffer_iter 10000
# JointMNIST - aux FMNIST
python experiments/train_ex_model.py --model lenet --scenario joint_mnist --strategy aux_data --lr 0.1 --epochs 100
######################
# ResNet #
######################
# CIFAR10, sampling-based cumulative
python experiments/train_ex_model.py --model resnet --scenario joint_cifar10 --strategy buffer_cumulative --buffer_size 500 --lr 0.1 --epochs 1000
python experiments/train_ex_model.py --model resnet --scenario joint_cifar10 --strategy minversion_cumulative --buffer_size 5000 --buffer_iter 10000 --buffer_tau 20 --buffer_wd 0.01 --lr 0.1
python experiments/train_ex_model.py --model resnet --scenario joint_cifar10 --strategy dimpression_cumulative --buffer_size 5000 --buffer_iter 10000 --buffer_tau 20 --buffer_wd 0.01 --lr 0.1
# CIFAR10, sampling-based naive
python experiments/train_ex_model.py --model resnet --scenario joint_cifar10 --strategy buffer_naive --buffer_size 500 --lr 0.1 --epochs 10000
python experiments/train_ex_model.py --model resnet --scenario joint_cifar10 --strategy dimpression_naive --buffer_size 5000 --buffer_iter 10000 --buffer_tau 20 --epochs 1000 --buffer_wd 0.01 --lr 0.1
python experiments/train_ex_model.py --model resnet --scenario joint_cifar10 --strategy minversion_naive --buffer_size 5000 --buffer_tau 20 --buffer_wd 0.01 --lr 0.1 --epochs 1000 --buffer_iter 10000
# SplitCIFAR100, sampling-based cumulative
python experiments/train_ex_model.py --model resnet --scenario split_cifar100 --strategy buffer_cumulative --buffer_size 500 --lr 0.1 --epochs 1000
python experiments/train_ex_model.py --model resnet --scenario split_cifar100 --strategy minversion_cumulative --buffer_size 500 --epochs 1000 --buffer_iter 10000 --buffer_tau 2 --epochs 1000 --buffer_wd 0.001 --lr 0.1
python experiments/train_ex_model.py --model resnet --scenario split_cifar100 --strategy dimpression_cumulative --buffer_size 500 --epochs 100 --buffer_iter 10000 --buffer_tau 20 --buffer_wd 0.01 --lr 0.1
python experiments/train_ex_model.py --model resnet --scenario split_cifar100 --strategy buffer_cumulative --buffer_size 500 --lr 0.1 --epochs 1000 --version class_mask
python experiments/train_ex_model.py --model resnet --scenario split_cifar100 --strategy minversion_cumulative --buffer_size 500 --epochs 1000 --buffer_iter 10000 --buffer_tau 2 --epochs 1000 --buffer_wd 0.001 --lr 0.1
python experiments/train_ex_model.py --model resnet --scenario split_cifar100 --strategy dimpression_cumulative --buffer_size 500 --epochs 1000 --buffer_iter 10000 --buffer_tau 2 --buffer_wd 0.001 --lr 0.1 --version v2