-
Notifications
You must be signed in to change notification settings - Fork 7
/
run_experiments.sh
241 lines (184 loc) · 11.9 KB
/
run_experiments.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
#!/bin/bash
########################
### MAIN EXPERIMENTS ###
########################
#################################################################
## NON-IMAGE ONLY ##
# export CUDA_VISIBLE_DEVICES=0
nohup python train.py --seed 0 --model non-image-only --n_TTA 0 --use_class_weights --label_smoothing 0.1 \
> non-image-only_ls0.1.out &
# export CUDA_VISIBLE_DEVICES=1
nohup python train.py --seed 1 --model non-image-only --n_TTA 0 --use_class_weights --label_smoothing 0.1 \
> non-image-only_ls0.1_seed1.out &
# export CUDA_VISIBLE_DEVICES=2
nohup python train.py --seed 2 --model non-image-only --n_TTA 0 --use_class_weights --label_smoothing 0.1 \
> non-image-only_ls0.1_seed2.out &
# export CUDA_VISIBLE_DEVICES=3
nohup python train.py --seed 3 --model non-image-only --n_TTA 0 --use_class_weights --label_smoothing 0.1 \
> non-image-only_ls0.1_seed3.out &
# export CUDA_VISIBLE_DEVICES=2
nohup python train.py --seed 4 --model non-image-only --n_TTA 0 --use_class_weights --label_smoothing 0.1 \
> non-image-only_ls0.1_seed4.out &
## END NON-IMAGE ONLY ##
#################################################################
#################################################################
## IMAGE ONLY ##
# export CUDA_VISIBLE_DEVICES=0
nohup python train.py --seed 0 --model image-only --n_TTA 5 --augment --use_class_weights --label_smoothing 0.1 \
> image-only_aug_ls0.1.out &
# export CUDA_VISIBLE_DEVICES=1
nohup python train.py --seed 1 --model image-only --n_TTA 5 --augment --use_class_weights --label_smoothing 0.1 \
> image-only_aug_ls0.1_seed1.out &
# export CUDA_VISIBLE_DEVICES=2
nohup python train.py --seed 2 --model image-only --n_TTA 5 --augment --use_class_weights --label_smoothing 0.1 \
> image-only_aug_ls0.1_seed2.out &
# export CUDA_VISIBLE_DEVICES=3
nohup python train.py --seed 3 --model image-only --n_TTA 5 --augment --use_class_weights --label_smoothing 0.1 \
> image-only_aug_ls0.1_seed3.out &
# export CUDA_VISIBLE_DEVICES=0
nohup python train.py --seed 4 --model image-only --n_TTA 5 --augment --use_class_weights --label_smoothing 0.1 \
> image-only_aug_ls0.1_seed4.out &
## END IMAGE ONLY ##
#################################################################
#################################################################
## PROBABILITY FUSION ##
# export CUDA_VISIBLE_DEVICES=0
nohup python train.py --seed 0 --model probability-fusion --n_TTA 5 --augment --use_class_weights --label_smoothing 0.1 \
> probability-fusion_aug_ls0.1.out &
# export CUDA_VISIBLE_DEVICES=1
nohup python train.py --seed 1 --model probability-fusion --n_TTA 5 --augment --use_class_weights --label_smoothing 0.1 \
> probability-fusion_aug_ls0.1_seed1.out &
# export CUDA_VISIBLE_DEVICES=2
nohup python train.py --seed 2 --model probability-fusion --n_TTA 5 --augment --use_class_weights --label_smoothing 0.1 \
> probability-fusion_aug_ls0.1_seed2.out &
# export CUDA_VISIBLE_DEVICES=3
nohup python train.py --seed 3 --model probability-fusion --n_TTA 5 --augment --use_class_weights --label_smoothing 0.1 \
> probability-fusion_aug_ls0.1_seed3.out &
# export CUDA_VISIBLE_DEVICES=1
nohup python train.py --seed 4 --model probability-fusion --n_TTA 5 --augment --use_class_weights --label_smoothing 0.1 \
> probability-fusion_aug_ls0.1_seed4.out &
## END PROBABILITY FUSION ##
#################################################################
#################################################################
## FEATURE FUSION ##
# export CUDA_VISIBLE_DEVICES=0
nohup python train.py --seed 0 --model feature-fusion --n_TTA 5 --augment --use_class_weights --label_smoothing 0.1 \
> feature-fusion_aug_ls0.1.out &
# export CUDA_VISIBLE_DEVICES=1
nohup python train.py --seed 1 --model feature-fusion --n_TTA 5 --augment --use_class_weights --label_smoothing 0.1 \
> feature-fusion_aug_ls0.1_seed1.out &
# export CUDA_VISIBLE_DEVICES=2
nohup python train.py --seed 2 --model feature-fusion --n_TTA 5 --augment --use_class_weights --label_smoothing 0.1 \
> feature-fusion_aug_ls0.1_seed2.out &
# export CUDA_VISIBLE_DEVICES=0
nohup python train.py --seed 3 --model feature-fusion --n_TTA 5 --augment --use_class_weights --label_smoothing 0.1 \
> feature-fusion_aug_ls0.1_seed3.out &
# export CUDA_VISIBLE_DEVICES=2
nohup python train.py --seed 4 --model feature-fusion --n_TTA 5 --augment --use_class_weights --label_smoothing 0.1 \
> feature-fusion_aug_ls0.1_seed4.out &
## END FEATURE FUSION ##
#################################################################
#################################################################
## LEARNED FEATURE FUSION (CONCAT) ##
# export CUDA_VISIBLE_DEVICES=0
nohup python train.py --seed 0 --model learned-feature-fusion --fusion_mode concat --n_TTA 5 --augment --use_class_weights --label_smoothing 0.1 \
> learned-feature-fusion_aug_ls0.1.out &
# export CUDA_VISIBLE_DEVICES=1
nohup python train.py --seed 1 --model learned-feature-fusion --fusion_mode concat --n_TTA 5 --augment --use_class_weights --label_smoothing 0.1 \
> learned-feature-fusion_aug_ls0.1_seed1.out &
# export CUDA_VISIBLE_DEVICES=2
nohup python train.py --seed 2 --model learned-feature-fusion --fusion_mode concat --n_TTA 5 --augment --use_class_weights --label_smoothing 0.1 \
> learned-feature-fusion_aug_ls0.1_seed2.out &
# export CUDA_VISIBLE_DEVICES=1
nohup python train.py --seed 3 --model learned-feature-fusion --fusion_mode concat --n_TTA 5 --augment --use_class_weights --label_smoothing 0.1 \
> learned-feature-fusion_aug_ls0.1_seed3.out &
# export CUDA_VISIBLE_DEVICES=2
nohup python train.py --seed 4 --model learned-feature-fusion --fusion_mode concat --n_TTA 5 --augment --use_class_weights --label_smoothing 0.1 \
> learned-feature-fusion_aug_ls0.1_seed4.out &
## END LEARNED FEATURE FUSION (CONCAT) ##
#################################################################
############################
### END MAIN EXPERIMENTS ###
############################
#############################
### AUXILIARY EXPERIMENTS ###
#############################
#################################################################
## LEARNED FEATURE FUSION (MULTIPLY) ##
# export CUDA_VISIBLE_DEVICES=0
nohup python train.py --seed 0 --model learned-feature-fusion --fusion_mode multiply --n_TTA 5 --augment --use_class_weights --label_smoothing 0.1 \
> learned-feature-fusion-multiply_aug_ls0.1.out &
# export CUDA_VISIBLE_DEVICES=1
nohup python train.py --seed 1 --model learned-feature-fusion --fusion_mode multiply --n_TTA 5 --augment --use_class_weights --label_smoothing 0.1 \
> learned-feature-fusion-multiply_aug_ls0.1_seed1.out &
# export CUDA_VISIBLE_DEVICES=2
nohup python train.py --seed 2 --model learned-feature-fusion --fusion_mode multiply --n_TTA 5 --augment --use_class_weights --label_smoothing 0.1 \
> learned-feature-fusion-multipy_aug_ls0.1_seed2.out &
# export CUDA_VISIBLE_DEVICES=1
nohup python train.py --seed 3 --model learned-feature-fusion --fusion_mode multiply --n_TTA 5 --augment --use_class_weights --label_smoothing 0.1 \
> learned-feature-fusion-multiply_aug_ls0.1_seed3.out &
# export CUDA_VISIBLE_DEVICES=2
nohup python train.py --seed 4 --model learned-feature-fusion --fusion_mode multiply --n_TTA 5 --augment --use_class_weights --label_smoothing 0.1 \
> learned-feature-fusion-multiply_aug_ls0.1_seed4.out &
## END LEARNED FEATURE FUSION (MULTIPLY) ##
#################################################################
#################################################################
## LEARNED FEATURE FUSION (ADD) ##
# export CUDA_VISIBLE_DEVICES=0
nohup python train.py --seed 0 --model learned-feature-fusion --fusion_mode add --n_TTA 5 --augment --use_class_weights --label_smoothing 0.1 \
> learned-feature-fusion_aug_ls0.1.out &
# export CUDA_VISIBLE_DEVICES=1
nohup python train.py --seed 1 --model learned-feature-fusion --fusion_mode add --n_TTA 5 --augment --use_class_weights --label_smoothing 0.1 \
> learned-feature-fusion_aug_ls0.1_seed1.out &
# export CUDA_VISIBLE_DEVICES=2
nohup python train.py --seed 2 --model learned-feature-fusion --fusion_mode add --n_TTA 5 --augment --use_class_weights --label_smoothing 0.1 \
> learned-feature-fusion_aug_ls0.1_seed2.out &
# export CUDA_VISIBLE_DEVICES=1
nohup python train.py --seed 3 --model learned-feature-fusion --fusion_mode add --n_TTA 5 --augment --use_class_weights --label_smoothing 0.1 \
> learned-feature-fusion_aug_ls0.1_seed3.out &
# export CUDA_VISIBLE_DEVICES=2
nohup python train.py --seed 4 --model learned-feature-fusion --fusion_mode add --n_TTA 5 --augment --use_class_weights --label_smoothing 0.1 \
> learned-feature-fusion_aug_ls0.1_seed4.out &
## END LEARNED FEATURE FUSION (ADD) ##
#################################################################
#################################################################
## LEARNED FEATURE FUSION VARIANT (EACH SUBNETWORK OPTIMIZED INDEPENDENTLY) ##
# export CUDA_VISIBLE_DEVICES=0
nohup python train.py --seed 0 --model learned-feature-fusion --train_mode multiopt --fusion_mode concat --n_TTA 5 --augment --use_class_weights --label_smoothing 0.1 \
> learned-feature-fusion_aug_ls0.1_multiopt.out &
# export CUDA_VISIBLE_DEVICES=1
nohup python train.py --seed 1 --model learned-feature-fusion --train_mode multiopt --fusion_mode concat --n_TTA 5 --augment --use_class_weights --label_smoothing 0.1 \
> learned-feature-fusion_aug_ls0.1_multiopt_seed1.out &
# export CUDA_VISIBLE_DEVICES=0
nohup python train.py --seed 2 --model learned-feature-fusion --train_mode multiopt --fusion_mode concat --n_TTA 5 --augment --use_class_weights --label_smoothing 0.1 \
> learned-feature-fusion_aug_ls0.1_multiopt_seed2.out &
# export CUDA_VISIBLE_DEVICES=3
nohup python train.py --seed 3 --model learned-feature-fusion --train_mode multiopt --fusion_mode concat --n_TTA 5 --augment --use_class_weights --label_smoothing 0.1 \
> learned-feature-fusion_aug_ls0.1_multiopt_seed3.out &
# export CUDA_VISIBLE_DEVICES=1
nohup python train.py --seed 4 --model learned-feature-fusion --train_mode multiopt --fusion_mode concat --n_TTA 5 --augment --use_class_weights --label_smoothing 0.1 \
> learned-feature-fusion_aug_ls0.1_multiopt_seed4.out &
## END LEARNED FEATURE FUSION VARIANT (EACH SUBNETWORK OPTIMIZED INDEPENDENTLY) ##
#################################################################
#################################################################
## LEARNED FEATURE FUSION VARIANT (TRAINED ON SUM OF THREE LOSSES) ##
# export CUDA_VISIBLE_DEVICES=1
nohup python train.py --seed 0 --model learned-feature-fusion --train_mode multiloss --fusion_mode concat --n_TTA 5 --augment --use_class_weights --label_smoothing 0.1 \
> learned-feature-fusion_aug_ls0.1_multiloss.out &
# export CUDA_VISIBLE_DEVICES=1
nohup python train.py --seed 1 --model learned-feature-fusion --train_mode multiloss --fusion_mode concat --n_TTA 5 --augment --use_class_weights --label_smoothing 0.1 \
> learned-feature-fusion_aug_ls0.1_multiloss_seed1.out &
# export CUDA_VISIBLE_DEVICES=2
nohup python train.py --seed 2 --model learned-feature-fusion --train_mode multiloss --fusion_mode concat --n_TTA 5 --augment --use_class_weights --label_smoothing 0.1 \
> learned-feature-fusion_aug_ls0.1_multiloss_seed2.out &
# export CUDA_VISIBLE_DEVICES=2
nohup python train.py --seed 3 --model learned-feature-fusion --train_mode multiloss --fusion_mode concat --n_TTA 5 --augment --use_class_weights --label_smoothing 0.1 \
> learned-feature-fusion_aug_ls0.1_multiloss_seed3.out &
# export CUDA_VISIBLE_DEVICES=3
nohup python train.py --seed 4 --model learned-feature-fusion --train_mode multiloss --fusion_mode concat --n_TTA 5 --augment --use_class_weights --label_smoothing 0.1 \
> learned-feature-fusion_aug_ls0.1_multiloss_seed4.out &
## LEARNED FEATURE FUSION (TRAINED ON SUM OF THREE LOSSES) ##
#################################################################
#################################
### END AUXILIARY EXPERIMENTS ###
#################################