-
Notifications
You must be signed in to change notification settings - Fork 0
/
run.sh
178 lines (111 loc) · 10.4 KB
/
run.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
#!/bin/bash
### current
#CUDA_VISIBLE_DEVICES=0,1,2,3 python train.py
#caliban run --experiment_config config.json train.py
#caliban cloud --name train_mlp_L1_W8_1024_S1_5 --xgroup train_mlp_L1_W8_1024_S1_5 --experiment_config config.json --gpu_spec 1xV100 train.py
#caliban cloud --name train_mlp_mnist_L1_W8_1024_S1_5 --xgroup train_mlp_mnist_L1_W8_1024_S1_5 --experiment_config config.json --gpu_spec 1xP100 train.py
#caliban cloud --name train_mlp_L1_W2048_32768_S1_5 --xgroup train_mlp_L1_W2048_32768_S1_5 --experiment_config config.json --gpu_spec 1xV100 train.py
#caliban cloud --name train_mlp_mnist_L1_W2048_32768_S1_5 --xgroup train_mlp_mnist_L1_W2048_32768_S1_5 --experiment_config config.json --gpu_spec 1xP100 train.py
#caliban cloud --name train_mlp_mnist_high --xgroup train_mlp_mnist_high --experiment_config config.json --gpu_spec 1xV100 train.py
#caliban cloud --name train_mlp_mnist_low1 --xgroup train_mlp_mnist_low1 --experiment_config config.json --gpu_spec 1xP100 train.py
#caliban cloud --name train_mlp_high --xgroup train_mlp_high --experiment_config config.json --gpu_spec 1xV100 train.py
#caliban cloud --name train_mlp_mnist_low2 --xgroup train_mlp_mnist_low2 --experiment_config config.json --gpu_spec 1xP100 train.py
#caliban cloud --name train_mlp_w4 --xgroup train_mlp_w4 --experiment_config config.json --gpu_spec 1xV100 train.py
#caliban run --experiment_config config.json train.py
#caliban cloud --name train_mlp_4K --xgroup train_mlp_4K --experiment_config config.json --gpu_spec 1xV100 train.py
#### width s_conv
#caliban run --experiment_config config.json train.py
#caliban cloud --name Sconv_width --xgroup Sconv_width --experiment_config config.json --gpu_spec 1xV100 train.py
##### width & Depth ResNet
#caliban run --experiment_config config.json train.py
#caliban cloud --name ResNet_widthDepth --xgroup ResNet_widthDepth --experiment_config config.json --gpu_spec 1xV100 train.py
##### width & Depth VGG
#caliban run --experiment_config config.json train.py
#caliban cloud --name VGG_width_depth --xgroup VGG_width_depth --experiment_config config.json --gpu_spec 1xV100 train.py
##### Depth
#caliban cloud --name depth_mlp_2layer --xgroup depth_mlp_2layer --experiment_config config.json --gpu_spec 1xV100 train.py
#caliban cloud --name depth_mlp_S1_6 --xgroup depth_mlp_S1_6 --experiment_config config.json --gpu_spec 1xV100 train.py
#caliban cloud --name depth_mlp --xgroup depth_mlp --experiment_config config.json --gpu_spec 1xV100 train.py
#caliban cloud --name depth_sconv --xgroup depth_sconv --experiment_config config.json --gpu_spec 1xV100 train.py
##### Barrier
#caliban run --experiment_config config_barrier.json barrier.py
#caliban cloud --name after_mlp_width_barrier --xgroup after_mlp_width_barrier --experiment_config config_barrier.json --gpu_spec 1xV100 barrier.py
#caliban cloud --name after_sconv_width_barrier --xgroup after_sconv_width_barrier --experiment_config config_barrier.json --gpu_spec 1xV100 barrier.py
#caliban cloud --name after_sconv_width_barrier_512_1024 --xgroup after_sconv_width_barrier_512_1024 --experiment_config config_barrier.json --gpu_spec 1xP100 barrier.py
#caliban cloud --name after_mlp_depth_barrier --xgroup after_mlp_depth_barrier --experiment_config config_barrier.json --gpu_spec 1xV100 barrier.py
#caliban cloud --name after_sconv_depth_barrier --xgroup after_sconv_depth_barrier --experiment_config config_barrier.json --gpu_spec 1xV100 barrier.py
#### barrier: resent and vgg
#CUDA_VISIBLE_DEVICES=0 python barrier.py
#caliban run --experiment_config config_barrier.json barrier.py
#caliban cloud --name before_vgg_barrier --xgroup before_vgg_barrier --experiment_config config_barrier.json --gpu_spec 1xV100 barrier.py
#caliban cloud --name before_resnet_barrier --xgroup before_resnet_barrier --experiment_config config_barrier.json --gpu_spec 1xV100 barrier.py
#### Basin Barrier
#caliban run --experiment_config config_barrier.json barrier_instance_v2.py
#caliban cloud --name basin_before_barrier --xgroup basin_before_barrier --experiment_config config_barrier.json --gpu_spec 1xV100 barrier_instance_v2.py
## after
#caliban cloud --name basin_mlp_width_barrier --xgroup basin_mlp_width_barrier --experiment_config config_barrier.json --gpu_spec 1xV100 barrier_instance_v2.py
#caliban cloud --name basin_sconv_width_barrier --xgroup basin_sconv_width_barrier --experiment_config config_barrier.json --gpu_spec 1xV100 barrier_instance_v2.py
#caliban cloud --name basin_mlp_depth_barrier --xgroup basin_mlp_depth_barrier --experiment_config config_barrier.json --gpu_spec 1xV100 barrier_instance_v2.py
##### FD
#CUDA_VISIBLE_DEVICES=0 python functional_diff_barrier_2layer.py
#caliban run --experiment_config config_barrier.json functional_diff_barrier_2layer.py
#caliban cloud --name mlp_width_barrier_FD --xgroup mlp_width_barrier_FD --experiment_config config_barrier.json --gpu_spec 1xP100 functional_diff_barrier_2layer.py
#### 1layer
#caliban run --experiment_config config_barrier.json functional_diff_barrier.py
#caliban cloud --name mlp_width_barrier_FD --xgroup mlp_width_barrier_FD --experiment_config config_barrier.json --gpu_spec 1xV100 functional_diff_barrier.py
##### Gaussian Noise
#CUDA_VISIBLE_DEVICES=0 python Ourmodel_GaussianPerturb.py
#CUDA_VISIBLE_DEVICES=0 python functional_diff_barrier_ourmodel.py
#caliban run --experiment_config config_barrier_FD.json functional_diff_barrier_ourmodel.py
##caliban cloud --name mlp_width_ourmodel_FD --xgroup mlp_width_ourmodel_FD --experiment_config config_barrier_FD.json --gpu_spec 1xV100 functional_diff_barrier_ourmodel.py
#caliban run --experiment_config config_barrier_FD.json FD_orthogonal_ourmodel.py
#caliban cloud --name mlp_width_ourmodel_FD --xgroup mlp_width_ourmodel_FD --experiment_config config_barrier_FD.json --gpu_spec 1xV100 FD_orthogonal_ourmodel.py
#### weight decay and Functional Difference
#caliban run --experiment_config config.json train.py
#caliban cloud --name train_mlp_WD_cifar10 --xgroup train_mlp_WD_cifar10 --experiment_config config.json --gpu_spec 1xV100 train.py
#caliban run --experiment_config config_barrier.json functional_diff_barrier.py
#### alpha our model and real world
#CUDA_VISIBLE_DEVICES=0 python functional_diff_barrier_alpha_ourmodel_realworld.py
### scaler Noise
#CUDA_VISIBLE_DEVICES=0 python functional_diff_barrier_ScalerNoise.py
#caliban run --experiment_config config_fd_fool.json functional_diff_barrier_ScalerNoise.py
#caliban cloud --name FD_fool_mlp --xgroup FD_fool_mlp --experiment_config config_fd_fool.json --gpu_spec 1xV100 functional_diff_barrier_ScalerNoise.py
#CUDA_VISIBLE_DEVICES=0 python functional_diff_barrier_GradientSimilarity.py
#CUDA_VISIBLE_DEVICES=0 python functional_diff_barrier_GradientSimilarity_hessian.py
#### deterministic training
#caliban run --experiment_config config_deterministic.json train_deterministic.py
#caliban cloud --name FD_fool_train --xgroup FD_fool_train --experiment_config config_deterministic.json --gpu_spec 1xV100 train_deterministic.py
#caliban run --experiment_config config_fd_fool_Gaussian_init.json functional_diff_barrier_Gaussian_init.py
#caliban cloud --name FD_fool_barrier --xgroup FD_fool_barrier --experiment_config config_fd_fool_Gaussian_init.json --gpu_spec 1xV100 functional_diff_barrier_Gaussian_init.py
### shared path
#caliban run --experiment_config config_sharedpath.json train_sharedpath.py
#caliban cloud --name FD_fool_train --xgroup FD_fool_train --experiment_config config_sharedpath.json --gpu_spec 1xV100 train_sharedpath.py
#caliban run --experiment_config config_fd_fool_shared_path.json functional_diff_barrier_SharedPath.py
#caliban cloud --name FD_fool_train --xgroup FD_fool_train --experiment_config config_fd_fool_shared_path.json --gpu_spec 1xV100 functional_diff_barrier_SharedPath.py
#### relaxed version of real world
#caliban run --experiment_config config_deterministic2.json train_deterministic2.py
#caliban cloud --name FD_fool_train --xgroup FD_fool_train --experiment_config config_deterministic2.json --gpu_spec 1xV100 train_deterministic2.py
#caliban run --experiment_config config_fd_fool_Gaussian_init.json functional_diff_barrier_Gaussian_init.py
#caliban cloud --name FD_fool_barrier --xgroup FD_fool_barrier --experiment_config config_fd_fool_Gaussian_init.json --gpu_spec 1xV100 functional_diff_barrier_Gaussian_init.py
### deterministic + scaling
#caliban run --experiment_config config_deterministic_scaled.json train_deterministic_scaled.py
#caliban cloud --name FD_train_scaled --xgroup FD_train_scaled --experiment_config config_deterministic_scaled.json --gpu_spec 1xV100 train_deterministic_scaled.py
#caliban run --experiment_config config_fd_fool_Gaussian_init_scaled.json functional_diff_barrier_Gaussian_init_scaled.py
#caliban cloud --name FD_barrier --xgroup FD_barrier --experiment_config config_fd_fool_Gaussian_init_scaled.json --gpu_spec 1xV100 functional_diff_barrier_Gaussian_init_scaled.py
######## FD scale
#### MLP 1layer
#caliban run --experiment_config config_fd_fool.json functional_diff_barrier_ScalerNoise_mlp1layer.py
#caliban cloud --name FD_fool_mlp1 --xgroup FD_fool_mlp1 --experiment_config config_fd_fool.json --gpu_spec 1xV100 functional_diff_barrier_ScalerNoise_mlp1layer.py
#caliban cloud --name FD_realworld_mlp1 --xgroup FD_realworld_mlp1 --experiment_config config_fd_fool.json --gpu_spec 1xV100 functional_diff_barrier_ScalerNoise_mlp1layer.py
### MLP 2layer
#caliban run --experiment_config config.json train.py
#caliban cloud --name mlp_depth_8k --xgroup mlp_depth_8k --experiment_config config.json --gpu_spec 1xV100 train.py
#caliban run --experiment_config config_fd_fool.json functional_diff_barrier_ScalerNoise.py
#caliban cloud --name FD_fool_mlp --xgroup FD_fool_mlp --experiment_config config_fd_fool.json --gpu_spec 1xV100 functional_diff_barrier_ScalerNoise.py
## 8K
#caliban cloud --name FD_fool_mlp --xgroup FD_fool_mlp --experiment_config config_fd_fool.json --gpu_spec 2xV100 functional_diff_barrier_ScalerNoise.py
### sconv-2layer
#CUDA_VISIBLE_DEVICES=0 python functional_diff_barrier_ScalerNoise_sconv2layer.py
#caliban run --experiment_config config_fd_fool.json functional_diff_barrier_ScalerNoise_sconv2layer.py
#caliban cloud --name FD_sconv2_realworld --xgroup FD_sconv2_realworld --experiment_config config_fd_fool.json --gpu_spec 2xV100 functional_diff_barrier_ScalerNoise_sconv2layer.py
caliban cloud --name FD_sconv2_ourmodel --xgroup FD_sconv2_ourmodel --experiment_config config_fd_fool.json --gpu_spec 2xV100 functional_diff_barrier_ScalerNoise_sconv2layer.py