-
Notifications
You must be signed in to change notification settings - Fork 0
Expand file tree
/
Copy pathscript.sh
More file actions
50 lines (34 loc) · 4.2 KB
/
Copy pathscript.sh
File metadata and controls
50 lines (34 loc) · 4.2 KB
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
## arguments: gpu, datast[coco, voc], batch-size, intermediate representation extracted at block of index[0, 1, 2]
#### intermediate representation extracted right after the first block
CUDA_VISIBLE_DEVICES=$1 python3 TL.py data/$2 --dataset=$2 --image-size 448 --batch-size $3 --seed=0 --model=resnet101 --optim_config=0 --epochs=30 --lr_scheduler --intermediate --where=$4
# CUDA_VISIBLE_DEVICES=$1 python3 TL.py data/$2 --dataset=$2 --image-size 384 --batch-size $3 --seed=0 --model=swin --optim_config=0 --epochs=30 --lr_scheduler --intermediate --where=$4
## EXAMPLES for Table1 and Table2a
#### resnet101 + intermediate representation extracted right after the second block
# CUDA_VISIBLE_DEVICES=0 python3 TL.py data/coco --dataset=coco --image-size 448 --batch-size 64 --seed=0 --model=resnet101 --optim_config=0 --epochs=30 --lr_scheduler --intermediate --where=1
#### resnet50 + intermediate representation extracted right after the third block
# CUDA_VISIBLE_DEVICES=0 python3 TL.py data/coco --dataset=coco --image-size 448 --batch-size 64 --seed=0 --model=resnet50 --optim_config=0 --epochs=30 --lr_scheduler --intermediate --where=2
#### swin + intermediate representation extracted right after the first block
# CUDA_VISIBLE_DEVICES=0 python3 TL.py data/coco --dataset=coco --image-size 384 --batch-size 50 --seed=0 --model=swin --optim_config=0 --epochs=30 --lr_scheduler --intermediate --where=0
#### swin large + intermediate representation extracted right after the second block
# CUDA_VISIBLE_DEVICES=0 python3 TL.py data/coco --dataset=coco --image-size 384 --batch-size 50 --seed=0 --model=swin_large --optim_config=0 --epochs=30 --lr_scheduler --intermediate --where=1
#### finetune
# CUDA_VISIBLE_DEVICES=0 python3 TL.py data/coco --dataset=coco --image-size 384 --batch-size 50 --seed=0 --model=swin_large --optim_config=0 --epochs=30 --lr_scheduler --finetune
#### Linear probe (baseline)
# CUDA_VISIBLE_DEVICES=0 python3 TL.py data/coco --dataset=coco --image-size 384 --batch-size 50 --seed=0 --model=swin_large --optim_config=1 --epochs=30 --lr_scheduler
## EXAMPLES for Table2b Pascal-VOC
#### resnet101 + intermediate representation extracted right after the second block
# CUDA_VISIBLE_DEVICES=0 python3 TL.py data/voc --dataset=voc --image-size 448 --batch-size 64 --seed=0 --model=resnet101 --optim_config=0 --epochs=30 --lr_scheduler --intermediate --where=1
#### resnet50 + intermediate representation extracted right after the third block
# CUDA_VISIBLE_DEVICES=0 python3 TL.py data/voc --dataset=voc --image-size 448 --batch-size 64 --seed=0 --model=resnet50 --optim_config=0 --epochs=30 --lr_scheduler --intermediate --where=2
#### swin + intermediate representation extracted right after the first block
# CUDA_VISIBLE_DEVICES=0 python3 TL.py data/voc --dataset=voc --image-size 384 --batch-size 50 --seed=0 --model=swin --optim_config=0 --epochs=30 --lr_scheduler --intermediate --where=0
#### swin large + intermediate representation extracted right after the second block
# CUDA_VISIBLE_DEVICES=0 python3 TL.py data/voc --dataset=voc --image-size 384 --batch-size 50 --seed=0 --model=swin_large --optim_config=0 --epochs=30 --lr_scheduler --intermediate --where=1
#### finetune
# CUDA_VISIBLE_DEVICES=0 python3 TL.py data/voc --dataset=voc --image-size 384 --batch-size 50 --seed=0 --model=swin_large --optim_config=0 --epochs=30 --lr_scheduler --finetune
#### Linear probe (baseline)
# CUDA_VISIBLE_DEVICES=0 python3 TL.py data/voc --dataset=voc --image-size 384 --batch-size 50 --seed=0 --model=swin_large --optim_config=1 --epochs=30 --lr_scheduler
## EXAMPLES for Appendix (Proposed method w/ fine-tune)
# CUDA_VISIBLE_DEVICES=0 python3 TL.py data/voc --dataset=voc --image-size 384 --batch-size 50 --seed=0 --model=resnet101 --optim_config=0 --epochs=30 --lr_scheduler --finetune --intermediate --where=1
# CUDA_VISIBLE_DEVICES=0 python3 TL.py data/voc --dataset=voc --image-size 384 --batch-size 50 --seed=0 --model=swin --optim_config=0 --epochs=30 --lr_scheduler --finetune --intermediate --where=1
# CUDA_VISIBLE_DEVICES=0 python3 TL.py data/coco --dataset=coco --image-size 384 --batch-size 50 --seed=0 --model=swin --optim_config=0 --epochs=30 --lr_scheduler --finetune --intermediate --where=1