Merge pull request #2 from SauravMaheshkar/main

Fix Weights and Biases Logging and paths

.gitignore

@@ -0,0 +1,4 @@
+**/__pycache__/
+*.pyc
+venv/
+wandb/

ease_of_teaching/forget_train.py

@@ -63,11 +63,15 @@
             group_name = args_raw.group_vars[0] + str(getattr(args_raw, args_raw.group_vars[0]))
             for var in args_raw.group_vars[1:]:
                 group_name = group_name + '_' + var + str(getattr(args_raw, var))
-        wandb.init(project="ease_of_teaching_forget",
-               group=args_raw.fname,
-               name=group_name)
+        wandb.init(project="fortuitous_forgetting",
+               group="ease_of_teaching")
         for var in args_raw.group_vars:
             wandb.config.update({var:getattr(args_raw, var)})
+    else:
+            wandb.init(project="fortuitous_forgetting",
+                       group="ease_of_teaching")
+            for var in args_raw.group_vars:
+                wandb.config.update({var: getattr(args_raw, var)})
 
 args = vars(args_raw) # convert python object to dict
 # args = parser.parse()  # parsed argument from CLI

llf_ke/configs/base_config.py

@@ -16,7 +16,7 @@ def __init__(self):
 
         # General Config
         parser.add_argument(
-            "--data", help="path to dataset base directory", default="/home/datasets"
+            "--data", help="path to dataset base directory", default="./datasets"
         )
 
         parser.add_argument("--optimizer", help="Which optimizer to use", default="sgd")

llf_ke/constants.py

@@ -1,2 +1,2 @@
-datasets_dir = '/home/datasets'
-checkpoints_dir = '/home/checkpoints'
+datasets_dir = './datasets'
+checkpoints_dir = './checkpoints'

llf_ke/train_KE_cls.py

@@ -235,11 +235,15 @@ def start_KE(cfg):
                 group_name = cfg.group_vars[0] + str(getattr(cfg, cfg.group_vars[0]))
                 for var in cfg.group_vars[1:]:
                     group_name = group_name + '_' + var + str(getattr(cfg, var))
-            wandb.init(project="llf_ke",
-                   group=cfg.group_name,
-                   name=group_name)
+            wandb.init(project="fortuitous_forgetting",
+                   group="llf_ke")
             for var in cfg.group_vars:
                 wandb.config.update({var:getattr(cfg, var)})
+        else:
+            wandb.init(project="fortuitous_forgetting",
+                       group="llf_ke")
+            for var in cfg.group_vars:
+                wandb.config.update({var: getattr(cfg, var)})
 
     if cfg.seed is not None and cfg.fix_seed: #FIXING SEED LEADS TO SAME REINITIALIZATION VALUES FOR EACH GENERATION
         random.seed(cfg.seed)

requirements.txt

@@ -5,7 +5,7 @@ pandas==1.3.5
 Pillow==9.0.0
 PyYAML==6.0
 scipy==1.7.3
-torch==1.7.0
-torchvision==0.11.2
+torch==1.10.2
+torchvision==0.11.3
 tqdm==4.62.3
 wandb==0.12.10

targeted_forgetting/mixed_group_training.py

@@ -21,7 +21,7 @@
 from torchvision.datasets.mnist import *
 ACT = F.relu
 
-data_path = '/home/datasets'
+data_path = './datasets'
 use_cuda = torch.cuda.is_available()
 
 
@@ -623,11 +623,16 @@ def main():
                 group_name = args.group_vars[0] + str(getattr(args, args.group_vars[0]))
                 for var in args.group_vars[1:]:
                     group_name = group_name + '_' + var + str(getattr(args, var))
-            wandb.init(project="mixed_group_training",
-                       group=args.fname,
+            wandb.init(project="fortuitous_forgetting",
+                       group="mixed_group_training",
                        name=group_name)
             for var in args.group_vars:
                 wandb.config.update({var: getattr(args, var)})
+        else:
+            wandb.init(project="fortuitous_forgetting",
+                       group="mixed_group_training")
+            for var in args.group_vars:
+                wandb.config.update({var: getattr(args, var)})
 
     if args.train_data == 'mnist':
         trans = ([transforms.ToTensor()])