update digits DA hyper-parameters

Author: JunguangJiang

examples/domain_adaptation/image_classification/README.md

@@ -19,6 +19,7 @@ pip install timm
 
 Following datasets can be downloaded automatically:
 
+- MNIST, SVHN, USPS 
 - Office31
 - OfficeCaltech
 - OfficeHome

examples/domain_adaptation/image_classification/cdan.sh

@@ -75,19 +75,19 @@ CUDA_VISIBLE_DEVICES=0 python cdan.py data/domainnet -d DomainNet -s c i p q s -
 CUDA_VISIBLE_DEVICES=0 python cdan.py data/domainnet -d DomainNet -s c i p q r -t s -a resnet101 --bottleneck-dim 1024 -r -rd 51200 --epochs 40 -i 5000 -p 500 --seed 0 --log logs/cdan/DomainNet_:2s
 
 # ResNet50, Wilds Dataset
-CUDA_VISIBLE_DEVICES=3 python cdan.py data/wilds -d iwildcam -a resnet50 --epochs 30 -i 1000 --seed 0 --log logs/cdan/iwildcam
+CUDA_VISIBLE_DEVICES=0 python cdan.py data/wilds -d iwildcam -a resnet50 --epochs 30 -i 1000 --seed 0 --log logs/cdan/iwildcam
 
 # DenseNet121, Wilds Dataset
-CUDA_VISIBLE_DEVICES=3 python cdan.py data/wilds -d camelyon17 --train-resizing 'res.' --val-resizing 'res.' --resize-size 96 \
+CUDA_VISIBLE_DEVICES=0 python cdan.py data/wilds -d camelyon17 --train-resizing 'res.' --val-resizing 'res.' --resize-size 96 \
   -a densenet121 --scratch --epochs 10 -i 1000 --lr 0.01 --seed 0 --log logs/cdan/camelyon17
-CUDA_VISIBLE_DEVICES=3 python cdan.py data/wilds -d fmow --train-resizing 'res.' --val-resizing 'res.' \
+CUDA_VISIBLE_DEVICES=0 python cdan.py data/wilds -d fmow --train-resizing 'res.' --val-resizing 'res.' \
   -a densenet121 --epochs 10 -i 1000 --lr 0.01 --seed 0 --log logs/cdan/fmow
 
 # Digits
-CUDA_VISIBLE_DEVICES=3 python cdan.py data/digits -d Digits -s MNIST -t USPS --train-resizing 'res.' --val-resizing 'res.' \
+CUDA_VISIBLE_DEVICES=0 python cdan.py data/digits -d Digits -s MNIST -t USPS --train-resizing 'res.' --val-resizing 'res.' \
   --resize-size 28 --no-hflip --norm-mean 0.5 --norm-std 0.5 -a lenet --no-pool --lr 0.01 -b 128 -i 2500 --scratch --seed 0 --log logs/cdan/MNIST2USPS
-CUDA_VISIBLE_DEVICES=3 python cdan.py data/digits -d Digits -s USPS -t MNIST --train-resizing 'res.' --val-resizing 'res.' \
+CUDA_VISIBLE_DEVICES=0 python cdan.py data/digits -d Digits -s USPS -t MNIST --train-resizing 'res.' --val-resizing 'res.' \
   --resize-size 28 --no-hflip --norm-mean 0.5 --norm-std 0.5 -a lenet --no-pool --lr 0.1 -b 128 -i 2500 --scratch --seed 0 --log logs/cdan/USPS2MNIST
-CUDA_VISIBLE_DEVICES=3 python cdan.py data/digits -d Digits -s SVHNRGB -t MNISTRGB --train-resizing 'res.' --val-resizing 'res.' \
-  --resize-size 32 --no-hflip --norm-mean 0.5 0.5 0.5 --norm-std 0.5 0.5 0.5 -a dtn --no-pool --lr 0.1 -b 128 -i 2500 --scratch --seed 0 --log logs/cdan/SVHN2MNIST
+CUDA_VISIBLE_DEVICES=0 python cdan.py data/digits -d Digits -s SVHNRGB -t MNISTRGB --train-resizing 'res.' --val-resizing 'res.' \
+  --resize-size 32 --no-hflip --norm-mean 0.5 0.5 0.5 --norm-std 0.5 0.5 0.5 -a dtn --no-pool --lr 0.01 -b 128 -i 2500 --scratch --seed 0 --log logs/cdan/SVHN2MNIST
 

examples/domain_adaptation/image_classification/dan.sh

@@ -74,19 +74,19 @@ CUDA_VISIBLE_DEVICES=0 python dan.py data/domainnet -d DomainNet -s c i p q s -t
 CUDA_VISIBLE_DEVICES=0 python dan.py data/domainnet -d DomainNet -s c i p q r -t s -a resnet101 --bottleneck-dim 1024 --epochs 40 -i 5000 -p 500 --seed 0 --log logs/dan/DomainNet_:2s
 
 # ResNet50, Wilds Dataset
-CUDA_VISIBLE_DEVICES=4 python dan.py data/wilds -d iwildcam -a resnet50 --epochs 30 -i 1000 --seed 0 --log logs/dan/iwildcam
+CUDA_VISIBLE_DEVICES=0 python dan.py data/wilds -d iwildcam -a resnet50 --epochs 30 -i 1000 --seed 0 --log logs/dan/iwildcam
 
 # DenseNet121, Wilds Dataset
-CUDA_VISIBLE_DEVICES=4 python dan.py data/wilds -d camelyon17 --train-resizing 'res.' --val-resizing 'res.' --resize-size 96 \
+CUDA_VISIBLE_DEVICES=0 python dan.py data/wilds -d camelyon17 --train-resizing 'res.' --val-resizing 'res.' --resize-size 96 \
   -a densenet121 --scratch --epochs 10 -i 1000 --lr 0.01 --seed 0 --log logs/dan/camelyon17
-CUDA_VISIBLE_DEVICES=4 python dan.py data/wilds -d fmow --train-resizing 'res.' --val-resizing 'res.' \
+CUDA_VISIBLE_DEVICES=0 python dan.py data/wilds -d fmow --train-resizing 'res.' --val-resizing 'res.' \
   -a densenet121 --epochs 10 -i 1000 --lr 0.01 --seed 0 --log logs/dan/fmow
 
 # Digits
-CUDA_VISIBLE_DEVICES=4 python dan.py data/digits -d Digits -s MNIST -t USPS --train-resizing 'res.' --val-resizing 'res.' \
+CUDA_VISIBLE_DEVICES=0 python dan.py data/digits -d Digits -s MNIST -t USPS --train-resizing 'res.' --val-resizing 'res.' \
   --resize-size 28 --no-hflip --norm-mean 0.5 --norm-std 0.5 -a lenet --no-pool --lr 0.01 -b 128 -i 2500 --scratch --seed 0 --log logs/dan/MNIST2USPS
-CUDA_VISIBLE_DEVICES=4 python dan.py data/digits -d Digits -s USPS -t MNIST --train-resizing 'res.' --val-resizing 'res.' \
+CUDA_VISIBLE_DEVICES=0 python dan.py data/digits -d Digits -s USPS -t MNIST --train-resizing 'res.' --val-resizing 'res.' \
   --resize-size 28 --no-hflip --norm-mean 0.5 --norm-std 0.5 -a lenet --no-pool --lr 0.1 -b 128 -i 2500 --scratch --seed 0 --log logs/dan/USPS2MNIST
-CUDA_VISIBLE_DEVICES=4 python dan.py data/digits -d Digits -s SVHNRGB -t MNISTRGB --train-resizing 'res.' --val-resizing 'res.' \
-  --resize-size 32 --no-hflip --norm-mean 0.5 0.5 0.5 --norm-std 0.5 0.5 0.5 -a dtn --no-pool --lr 0.1 -b 128 -i 2500 --scratch --seed 0 --log logs/dan/SVHN2MNIST
+CUDA_VISIBLE_DEVICES=0 python dan.py data/digits -d Digits -s SVHNRGB -t MNISTRGB --train-resizing 'res.' --val-resizing 'res.' \
+  --resize-size 32 --no-hflip --norm-mean 0.5 0.5 0.5 --norm-std 0.5 0.5 0.5 -a dtn --no-pool --lr 0.01 -b 128 -i 2500 --scratch --seed 0 --log logs/dan/SVHN2MNIST
 

examples/domain_adaptation/image_classification/dann.sh

@@ -74,19 +74,18 @@ CUDA_VISIBLE_DEVICES=0 python dann.py data/domainnet -d DomainNet -s c i p q s -
 CUDA_VISIBLE_DEVICES=0 python dann.py data/domainnet -d DomainNet -s c i p q r -t s -a resnet101 --bottleneck-dim 1024 --epochs 40 -i 5000 -p 500 --seed 0 --log logs/dann/DomainNet_:2s
 
 # ResNet50, Wilds Dataset
-CUDA_VISIBLE_DEVICES=5 python dann.py data/wilds -d iwildcam -a resnet50 --epochs 30 -i 1000 --seed 0 --log logs/dann/iwildcam
+CUDA_VISIBLE_DEVICES=0 python dann.py data/wilds -d iwildcam -a resnet50 --epochs 30 -i 1000 --seed 0 --log logs/dann/iwildcam
 
 # DenseNet121, Wilds Dataset
-CUDA_VISIBLE_DEVICES=5 python dann.py data/wilds -d camelyon17 --train-resizing 'res.' --val-resizing 'res.' --resize-size 96 \
+CUDA_VISIBLE_DEVICES=0 python dann.py data/wilds -d camelyon17 --train-resizing 'res.' --val-resizing 'res.' --resize-size 96 \
   -a densenet121 --scratch --epochs 10 -i 1000 --lr 0.01 --seed 0 --log logs/dann/camelyon17
-CUDA_VISIBLE_DEVICES=5 python dann.py data/wilds -d fmow --train-resizing 'res.' --val-resizing 'res.' \
+CUDA_VISIBLE_DEVICES=0 python dann.py data/wilds -d fmow --train-resizing 'res.' --val-resizing 'res.' \
   -a densenet121 --epochs 10 -i 1000 --lr 0.01 --seed 0 --log logs/dann/fmow
 
 # Digits
-CUDA_VISIBLE_DEVICES=5 python dann.py data/digits -d Digits -s MNIST -t USPS --train-resizing 'res.' --val-resizing 'res.' \
+CUDA_VISIBLE_DEVICES=0 python dann.py data/digits -d Digits -s MNIST -t USPS --train-resizing 'res.' --val-resizing 'res.' \
   --resize-size 28 --no-hflip --norm-mean 0.5 --norm-std 0.5 -a lenet --no-pool --lr 0.01 -b 128 -i 2500 --scratch --seed 0 --log logs/dann/MNIST2USPS
-CUDA_VISIBLE_DEVICES=5 python dann.py data/digits -d Digits -s USPS -t MNIST --train-resizing 'res.' --val-resizing 'res.' \
+CUDA_VISIBLE_DEVICES=0 python dann.py data/digits -d Digits -s USPS -t MNIST --train-resizing 'res.' --val-resizing 'res.' \
   --resize-size 28 --no-hflip --norm-mean 0.5 --norm-std 0.5 -a lenet --no-pool --lr 0.1 -b 128 -i 2500 --scratch --seed 0 --log logs/dann/USPS2MNIST
-CUDA_VISIBLE_DEVICES=5 python dann.py data/digits -d Digits -s SVHNRGB -t MNISTRGB --train-resizing 'res.' --val-resizing 'res.' \
+CUDA_VISIBLE_DEVICES=0 python dann.py data/digits -d Digits -s SVHNRGB -t MNISTRGB --train-resizing 'res.' --val-resizing 'res.' \
   --resize-size 32 --no-hflip --norm-mean 0.5 0.5 0.5 --norm-std 0.5 0.5 0.5 -a dtn --no-pool --lr 0.1 -b 128 -i 2500 --scratch --seed 0 --log logs/dann/SVHN2MNIST
-

examples/domain_adaptation/image_classification/jan.sh

@@ -74,19 +74,19 @@ CUDA_VISIBLE_DEVICES=0 python jan.py data/domainnet -d DomainNet -s c i p q s -t
 CUDA_VISIBLE_DEVICES=0 python jan.py data/domainnet -d DomainNet -s c i p q r -t s -a resnet101 --bottleneck-dim 1024 --epochs 40 -i 5000 -p 500 --seed 0 --log logs/jan/DomainNet_:2s
 
 # ResNet50, Wilds Dataset
-CUDA_VISIBLE_DEVICES=6 python jan.py data/wilds -d iwildcam -a resnet50 --epochs 30 -i 1000 --seed 0 --log logs/jan/iwildcam
+CUDA_VISIBLE_DEVICES=0 python jan.py data/wilds -d iwildcam -a resnet50 --epochs 30 -i 1000 --seed 0 --log logs/jan/iwildcam
 
 # DenseNet121, Wilds Dataset
-CUDA_VISIBLE_DEVICES=6 python jan.py data/wilds -d camelyon17 --train-resizing 'res.' --val-resizing 'res.' --resize-size 96 \
+CUDA_VISIBLE_DEVICES=0 python jan.py data/wilds -d camelyon17 --train-resizing 'res.' --val-resizing 'res.' --resize-size 96 \
   -a densenet121 --scratch --epochs 10 -i 1000 --lr 0.01 --seed 0 --log logs/jan/camelyon17
-CUDA_VISIBLE_DEVICES=6 python jan.py data/wilds -d fmow --train-resizing 'res.' --val-resizing 'res.' \
+CUDA_VISIBLE_DEVICES=0 python jan.py data/wilds -d fmow --train-resizing 'res.' --val-resizing 'res.' \
   -a densenet121 --epochs 10 -i 1000 --lr 0.01 --seed 0 --log logs/jan/fmow
 
 # Digits
-CUDA_VISIBLE_DEVICES=6 python jan.py data/digits -d Digits -s MNIST -t USPS --train-resizing 'res.' --val-resizing 'res.' \
+CUDA_VISIBLE_DEVICES=0 python jan.py data/digits -d Digits -s MNIST -t USPS --train-resizing 'res.' --val-resizing 'res.' \
   --resize-size 28 --no-hflip --norm-mean 0.5 --norm-std 0.5 -a lenet --no-pool --lr 0.01 -b 128 -i 2500 --scratch --seed 0 --log logs/jan/MNIST2USPS
-CUDA_VISIBLE_DEVICES=6 python jan.py data/digits -d Digits -s USPS -t MNIST --train-resizing 'res.' --val-resizing 'res.' \
+CUDA_VISIBLE_DEVICES=0 python jan.py data/digits -d Digits -s USPS -t MNIST --train-resizing 'res.' --val-resizing 'res.' \
   --resize-size 28 --no-hflip --norm-mean 0.5 --norm-std 0.5 -a lenet --no-pool --lr 0.1 -b 128 -i 2500 --scratch --seed 0 --log logs/jan/USPS2MNIST
-CUDA_VISIBLE_DEVICES=6 python jan.py data/digits -d Digits -s SVHNRGB -t MNISTRGB --train-resizing 'res.' --val-resizing 'res.' \
+CUDA_VISIBLE_DEVICES=0 python jan.py data/digits -d Digits -s SVHNRGB -t MNISTRGB --train-resizing 'res.' --val-resizing 'res.' \
   --resize-size 32 --no-hflip --norm-mean 0.5 0.5 0.5 --norm-std 0.5 0.5 0.5 -a dtn --no-pool --lr 0.1 -b 128 -i 2500 --scratch --seed 0 --log logs/jan/SVHN2MNIST
 

examples/domain_adaptation/image_classification/mcc.sh

@@ -74,19 +74,19 @@ CUDA_VISIBLE_DEVICES=0 python mcc.py data/domainnet -d DomainNet -s c i p q s -t
 CUDA_VISIBLE_DEVICES=0 python mcc.py data/domainnet -d DomainNet -s c i p q r -t s -a resnet101 --bottleneck-dim 2048 --epochs 40 -i 5000 -p 500 --seed 0 --log logs/mcc/DomainNet_:2s
 
 # ResNet50, Wilds Dataset
-CUDA_VISIBLE_DEVICES=7 python mcc.py data/wilds -d iwildcam -a resnet50 --epochs 30 -i 1000 --seed 0 --bottleneck-dim 2048 --log logs/mcc/iwildcam
+CUDA_VISIBLE_DEVICES=0 python mcc.py data/wilds -d iwildcam -a resnet50 --epochs 30 -i 1000 --seed 0 --bottleneck-dim 2048 --log logs/mcc/iwildcam
 
 # DenseNet121, Wilds Dataset
-CUDA_VISIBLE_DEVICES=7 python mcc.py data/wilds -d camelyon17 --train-resizing 'res.' --val-resizing 'res.' --resize-size 96 \
+CUDA_VISIBLE_DEVICES=0 python mcc.py data/wilds -d camelyon17 --train-resizing 'res.' --val-resizing 'res.' --resize-size 96 \
   -a densenet121 --scratch --epochs 10 -i 1000 --lr 0.01 --seed 0 --log logs/mcc/camelyon17
-CUDA_VISIBLE_DEVICES=7 python mcc.py data/wilds -d fmow --train-resizing 'res.' --val-resizing 'res.' \
+CUDA_VISIBLE_DEVICES=0 python mcc.py data/wilds -d fmow --train-resizing 'res.' --val-resizing 'res.' \
   -a densenet121 --epochs 10 -i 1000 --lr 0.01 --seed 0 --bottleneck-dim 2048 --log logs/mcc/fmow
 
 # Digits
-CUDA_VISIBLE_DEVICES=7 python mcc.py data/digits -d Digits -s MNIST -t USPS --train-resizing 'res.' --val-resizing 'res.' \
+CUDA_VISIBLE_DEVICES=0 python mcc.py data/digits -d Digits -s MNIST -t USPS --train-resizing 'res.' --val-resizing 'res.' \
   --resize-size 28 --no-hflip --norm-mean 0.5 --norm-std 0.5 -a lenet --no-pool --lr 0.01 -b 128 -i 2500 --scratch --seed 0 --log logs/mcc/MNIST2USPS
-CUDA_VISIBLE_DEVICES=7 python mcc.py data/digits -d Digits -s USPS -t MNIST --train-resizing 'res.' --val-resizing 'res.' \
+CUDA_VISIBLE_DEVICES=0 python mcc.py data/digits -d Digits -s USPS -t MNIST --train-resizing 'res.' --val-resizing 'res.' \
   --resize-size 28 --no-hflip --norm-mean 0.5 --norm-std 0.5 -a lenet --no-pool --lr 0.1 -b 128 -i 2500 --scratch --seed 0 --log logs/mcc/USPS2MNIST
-CUDA_VISIBLE_DEVICES=7 python mcc.py data/digits -d Digits -s SVHNRGB -t MNISTRGB --train-resizing 'res.' --val-resizing 'res.' \
-  --resize-size 32 --no-hflip --norm-mean 0.5 0.5 0.5 --norm-std 0.5 0.5 0.5 -a dtn --no-pool --lr 0.1 -b 128 -i 2500 --scratch --seed 0 --log logs/mcc/SVHN2MNIST
+CUDA_VISIBLE_DEVICES=0 python mcc.py data/digits -d Digits -s SVHNRGB -t MNISTRGB --train-resizing 'res.' --val-resizing 'res.' \
+  --resize-size 32 --no-hflip --norm-mean 0.5 0.5 0.5 --norm-std 0.5 0.5 0.5 -a dtn --no-pool --lr 0.01 -b 128 -i 2500 --scratch --seed 0 --log logs/mcc/SVHN2MNIST
 

examples/domain_adaptation/image_classification/mdd.sh

@@ -7,20 +7,6 @@ CUDA_VISIBLE_DEVICES=0 python mdd.py data/office31 -d Office31 -s A -t D -a resn
 CUDA_VISIBLE_DEVICES=0 python mdd.py data/office31 -d Office31 -s D -t A -a resnet50 --epochs 20 --bottleneck-dim 1024 --seed 1 --log logs/mdd/Office31_D2A
 CUDA_VISIBLE_DEVICES=0 python mdd.py data/office31 -d Office31 -s W -t A -a resnet50 --epochs 20 --bottleneck-dim 1024 --seed 1 --log logs/mdd/Office31_W2A
 
-CUDA_VISIBLE_DEVICES=0 python mdd.py data/office31 -d Office31 -s A -t W -a resnet50 --epochs 20 --bottleneck-dim 1024 --seed 2 --log logs/mdd/Office31_A2W
-CUDA_VISIBLE_DEVICES=0 python mdd.py data/office31 -d Office31 -s D -t W -a resnet50 --epochs 20 --bottleneck-dim 1024 --seed 2 --log logs/mdd/Office31_D2W
-CUDA_VISIBLE_DEVICES=0 python mdd.py data/office31 -d Office31 -s W -t D -a resnet50 --epochs 20 --bottleneck-dim 1024 --seed 2 --log logs/mdd/Office31_W2D
-CUDA_VISIBLE_DEVICES=0 python mdd.py data/office31 -d Office31 -s A -t D -a resnet50 --epochs 20 --bottleneck-dim 1024 --seed 2 --log logs/mdd/Office31_A2D
-CUDA_VISIBLE_DEVICES=0 python mdd.py data/office31 -d Office31 -s D -t A -a resnet50 --epochs 20 --bottleneck-dim 1024 --seed 2 --log logs/mdd/Office31_D2A
-CUDA_VISIBLE_DEVICES=0 python mdd.py data/office31 -d Office31 -s W -t A -a resnet50 --epochs 20 --bottleneck-dim 1024 --seed 2 --log logs/mdd/Office31_W2A
-
-CUDA_VISIBLE_DEVICES=0 python mdd.py data/office31 -d Office31 -s A -t W -a resnet50 --epochs 20 --bottleneck-dim 1024 --seed 3 --log logs/mdd/Office31_A2W
-CUDA_VISIBLE_DEVICES=0 python mdd.py data/office31 -d Office31 -s D -t W -a resnet50 --epochs 20 --bottleneck-dim 1024 --seed 3 --log logs/mdd/Office31_D2W
-CUDA_VISIBLE_DEVICES=0 python mdd.py data/office31 -d Office31 -s W -t D -a resnet50 --epochs 20 --bottleneck-dim 1024 --seed 3 --log logs/mdd/Office31_W2D
-CUDA_VISIBLE_DEVICES=0 python mdd.py data/office31 -d Office31 -s A -t D -a resnet50 --epochs 20 --bottleneck-dim 1024 --seed 3 --log logs/mdd/Office31_A2D
-CUDA_VISIBLE_DEVICES=0 python mdd.py data/office31 -d Office31 -s D -t A -a resnet50 --epochs 20 --bottleneck-dim 1024 --seed 3 --log logs/mdd/Office31_D2A
-CUDA_VISIBLE_DEVICES=0 python mdd.py data/office31 -d Office31 -s W -t A -a resnet50 --epochs 20 --bottleneck-dim 1024 --seed 3 --log logs/mdd/Office31_W2A
-
 # ResNet50, Office-Home, Single Source
 CUDA_VISIBLE_DEVICES=0 python mdd.py data/office-home -d OfficeHome -s Ar -t Cl -a resnet50 --epochs 30 --bottleneck-dim 2048 --seed 0 --log logs/mdd/OfficeHome_Ar2Cl
 CUDA_VISIBLE_DEVICES=0 python mdd.py data/office-home -d OfficeHome -s Ar -t Pr -a resnet50 --epochs 30 --bottleneck-dim 2048 --seed 0 --log logs/mdd/OfficeHome_Ar2Pr
@@ -91,19 +77,19 @@ CUDA_VISIBLE_DEVICES=0 python mdd.py data/domainnet -d DomainNet -s c i p q s -t
 CUDA_VISIBLE_DEVICES=0 python mdd.py data/domainnet -d DomainNet -s c i p q r -t s -a resnet101 --epochs 40 -i 5000 -p 500 --bottleneck-dim 2048 --seed 0 --lr 0.004 --log logs/mdd/DomainNet_:2s
 
 # ResNet50, Wilds Dataset
-CUDA_VISIBLE_DEVICES=0 python mdd.py data/wilds -d iwildcam --train-resizing 'res' --val-resizing 'res' \
+CUDA_VISIBLE_DEVICES=0 python mdd.py data/wilds -d iwildcam --train-resizing 'res.' --val-resizing 'res.' \
   -a resnet50 --bottleneck-dim 2048 --epochs 30 -i 1000 --seed 0 --log logs/mdd/iwildcam
 
 # DenseNet121, Wilds Dataset
-CUDA_VISIBLE_DEVICES=0 python mdd.py data/wilds -d camelyon17 --train-resizing 'res' --val-resizing 'res' --resize-size 96 \
+CUDA_VISIBLE_DEVICES=0 python mdd.py data/wilds -d camelyon17 --train-resizing 'res.' --val-resizing 'res.' --resize-size 96 \
   -a densenet121 --scratch --epochs 30 -i 1000 --lr 0.01 --seed 0 --log logs/mdd/camelyon17
-CUDA_VISIBLE_DEVICES=0 python mdd.py data/wilds -d fmow --train-resizing 'res' --val-resizing 'res' \
+CUDA_VISIBLE_DEVICES=0 python mdd.py data/wilds -d fmow --train-resizing 'res.' --val-resizing 'res.' \
   -a densenet121 --bottleneck-dim 2048 --epochs 30 -i 1000 --lr 0.01 --seed 0 --log logs/mdd/fmow
 
 # Digits
-CUDA_VISIBLE_DEVICES=1 python mdd.py data/digits -d Digits -s MNIST -t USPS --train-resizing 'res' --val-resizing 'res' \
+CUDA_VISIBLE_DEVICES=0 python mdd.py data/digits -d Digits -s MNIST -t USPS --train-resizing 'res.' --val-resizing 'res.' \
   --resize-size 28 --no-hflip --norm-mean 0.5 --norm-std 0.5 -a lenet --no-pool --lr 0.01 -b 128 -i 2500 --scratch --seed 0 --log logs/mdd/MNIST2USPS
-CUDA_VISIBLE_DEVICES=1 python mdd.py data/digits -d Digits -s USPS -t MNIST --train-resizing 'res' --val-resizing 'res' \
+CUDA_VISIBLE_DEVICES=0 python mdd.py data/digits -d Digits -s USPS -t MNIST --train-resizing 'res.' --val-resizing 'res.' \
   --resize-size 28 --no-hflip --norm-mean 0.5 --norm-std 0.5 -a lenet --no-pool --lr 0.01 -b 128 -i 2500 --scratch --seed 0 --log logs/mdd/USPS2MNIST
-CUDA_VISIBLE_DEVICES=1 python mdd.py data/digits -d Digits -s SVHNRGB -t MNISTRGB --train-resizing 'res' --val-resizing 'res' \
+CUDA_VISIBLE_DEVICES=0 python mdd.py data/digits -d Digits -s SVHNRGB -t MNISTRGB --train-resizing 'res.' --val-resizing 'res.' \
   --resize-size 32 --no-hflip --norm-mean 0.5 0.5 0.5 --norm-std 0.5 0.5 0.5 -a dtn --no-pool --lr 0.01 -b 128 -i 2500 --scratch --seed 0 --log logs/mdd/SVHN2MNIST

examples/domain_adaptation/image_classification/self_ensemble.sh

@@ -25,10 +25,6 @@ CUDA_VISIBLE_DEVICES=0 python self_ensemble.py data/office-home -d OfficeHome -s
 CUDA_VISIBLE_DEVICES=0 python self_ensemble.py data/visda-2017 -d VisDA2017 -s Synthetic -t Real -a resnet101 \
     --epochs 20 --seed 0 --per-class-eval --log logs/self_ensemble/VisDA2017 --lr-gamma 0.0002 -b 32
 
-# Wilds Dataset
-CUDA_VISIBLE_DEVICES=0 python self_ensemble.py data/wilds -d iwildcam -a resnet50 --epochs 30 -i 1000 --seed 0 --log logs/self_ensemble/iwildcam
-CUDA_VISIBLE_DEVICES=0 python self_ensemble.py data/wilds -d fmow -a resnet50 --epochs 30 -i 1000 --seed 0 --log logs/self_ensemble/fmow
-
 # Office-Home on Vision Transformer
 CUDA_VISIBLE_DEVICES=0 python self_ensemble.py data/office-home -d OfficeHome -s Ar -t Cl -a vit_base_patch16_224 --no-pool --epochs 30 --seed 0 -b 24 --log logs/self_ensemble_vit/OfficeHome_Ar2Cl
 CUDA_VISIBLE_DEVICES=0 python self_ensemble.py data/office-home -d OfficeHome -s Ar -t Pr -a vit_base_patch16_224 --no-pool --epochs 30 --seed 0 -b 24 --log logs/self_ensemble_vit/OfficeHome_Ar2Pr