new

bash_files/linear/cifar10/simclr.sh → bash_files/linear/cifar/simclr.sh

@@ -16,5 +16,8 @@ python3 ../../../main_linear.py \
     --name $1-linear \
     --entity doxawang \
     --project solo-learn \
-    $4 \
-    --wandb 
+    --pretrained_feature_extractor $4 \
+    --use_poison \
+    --poison_data $5 \
+    --load_linear \
+    # --wandb    

bash_files/pretrain/cifar/exe.py

@@ -1,15 +1,16 @@
 rate = '0.60'
 dataset = 'cifar10'
-# poison_method = 'zoo-simclr'
-poison_method = 'clb'
+poison_method = 'zoo-simclr'
+# poison_method = 'clb'
 import time
 out_dir = '/data/yfwang/solo-learn-outputs'
 
 def sweep_poison_rate(args):
     i = 0
     # for dataset in ['cifar10', 'cifar100']:
-    for dataset in ['cifar100']:
-        for rate in '0.10 0.20 0.30 0.40 0.50 0.60 0.70 0.80 0.90 1.00'.split(' '):
+    for dataset in ['cifar10']:
+        # for rate in '0.10 0.20 0.30 0.40 0.50 0.60 0.70 0.80 0.90 1.00'.split(' '):
+        for rate in '0.90 1.00 1.10 1.20'.split(' '):
             gpu = args.gpus[i]
             print(rate)
             os.system(f"""
@@ -36,8 +37,8 @@ def sweep_pretrain_method(args):
         os.system(f"""
         for file in {out_dir}/poison_datasets/{dataset}/{poison_method}/gaussian_noise/{dataset}_{poison_method}_rate_{rate}_*.pt
         do 
-        # echo ${{file}}, {method}
-        CUDA_VISIBLE_DEVICES={gpu} sh {method}.sh {dataset} " --poison_data ${{file}}  --use_poison --checkpoint_dir {out_dir}/pretrain/{dataset} " &
+        # echo ${{file}}, {method}, {dataset}, {out_dir}
+        # CUDA_VISIBLE_DEVICES={gpu} sh {method}.sh {dataset} " --poison_data ${{file}}  --use_poison --checkpoint_dir {out_dir}/pretrain/{dataset} " &
         done
         """
         )
@@ -97,6 +98,93 @@ def sweep_cifar100(args):
                         return
                     time.sleep(1)
 
+
+def sweep_poison_pretrain_method(args):
+    i = 0
+    # for dataset in ['cifar10', 'cifar100']:
+    # for method in 'dino'.split(' '):
+    for method in ['supcon', 'mocov2plus', 'byol', 'simsiam', 'swav', 'dino', 'barlow']:
+        poison_method = 'zoo-' + method
+        # for rate in '0.10 0.20 0.30 0.40 0.50 0.60 0.70 0.80 0.90 1.00'.split(' '):
+        gpu = args.gpus[i]
+        print(rate)
+
+        os.system(f"""
+        for file in {out_dir}/poison_datasets/{dataset}/{poison_method}/gaussian_noise/{dataset}_{poison_method}_rate_{rate}_*.pt
+        do 
+        echo ${{file}}, {method}
+        # CUDA_VISIBLE_DEVICES={gpu} sh {method}.sh {dataset} " --poison_data ${{file}}  --use_poison --checkpoint_dir {out_dir}/pretrain/{dataset} " &
+        CUDA_VISIBLE_DEVICES={gpu} sh simclr.sh {dataset} " --poison_data ${{file}}  --use_poison --checkpoint_dir {out_dir}/pretrain/{dataset} " &
+        done
+        """
+        )
+        i += 1
+        if i >= len(args.gpus):
+            return
+        time.sleep(1)
+
+def sweep_alpha(args):
+    i = 0
+    for dataset in ['cifar10', 'cifar100']:
+        # for alpha in '0.05 0.10'.split(' '):
+        for alpha in '0.15'.split(' '):
+            gpu = args.gpus[i]
+            print(rate)
+            os.system(f"""
+            for file in /data/yfwang/solo-learn/poison_datasets/{dataset}/zoo-simclr/gaussian_noise/{dataset}_zoo-simclr_rate_{rate}_*_{alpha}*.pt
+            do 
+            # echo ${{file}}
+            CUDA_VISIBLE_DEVICES={gpu} sh simclr.sh {dataset} " --poison_data ${{file}}  --use_poison --checkpoint_dir /data/yfwang/solo-learn/pretrain/{dataset} " &
+            done
+            """
+            )
+            i += 1
+            if i >= len(args.gpus):
+                return
+
+def sweep_trigger(args):
+    i = 0
+    for dataset in ['cifar10', 'cifar100']:
+        # for alpha in '0.05 0.10'.split(' '):
+        # for alpha in '0.15'.split(' '):
+        for trigger in 'checkerboard_1corner checkerboard_4corner checkerboard_center checkerboard_full gaussian_noise'.split(' '):
+            gpu = args.gpus[i]
+            print(rate)
+            os.system(f"""
+            for file in {out_dir}/poison_datasets/{dataset}/zoo-simclr/{trigger}/{dataset}_zoo-simclr_rate_{rate}_target_None_trigger_{trigger}_*.pt
+            do 
+            # rm $file
+            # echo ${{file}}
+            CUDA_VISIBLE_DEVICES={gpu} sh simclr.sh {dataset} " --poison_data ${{file}}  --use_poison --checkpoint_dir /data/yfwang/solo-learn/pretrain/{dataset} " &
+            done
+            """
+            )
+            i += 1
+            if i >= len(args.gpus):
+                return
+
+def sweep_trial(args):
+    i = 0
+    # for dataset in ['cifar10', 'cifar100']:
+        # for alpha in '0.05 0.10'.split(' '):
+        # for alpha in '0.15'.split(' '):
+    for budget in '5 50'.split(' '):
+        for trial in '0 1 2 3 4'.split(' '):
+            gpu = args.gpus[i]
+            print(rate)
+            os.system(f"""
+            for file in {out_dir}/poison_datasets/{dataset}/{poison_method}/gaussian_noise/budget_{budget}/{dataset}_{poison_method}_rate_{rate}_target_0_trigger_gaussian_noise_alpha_0.20_class_0_trial_{trial}_*.pt
+            do 
+            # rm $file
+            # echo ${{file}}
+            CUDA_VISIBLE_DEVICES={gpu} sh simclr.sh {dataset} " --poison_data ${{file}}  --use_poison --checkpoint_dir /data/yfwang/solo-learn/pretrain/{dataset} " &
+            done
+            """
+            )
+            i += 1
+            if i >= len(args.gpus):
+                return
+
 if __name__ == "__main__":
     import argparse
     import os 
@@ -108,5 +196,10 @@ def sweep_cifar100(args):
     args = parser.parse_args()
     # sweep_pretrain_method(args)
     # sweep_pretrain_method(args)
-    # sweep_eval(args)/
-    sweep_cifar100(args)
+    # sweep_eval(args)
+    # sweep_cifar100(args)
+    # sweep_poison_pretrain_method(args)
+    # sweep_alpha(args)
+    # sweep_poison_rate(args)
+    # sweep_trigger(args)
+    sweep_trial(args)