disable random resize crop + ft config

configs/supervised_wilds_vith14_ep300-ft.yaml

@@ -0,0 +1,49 @@
+meta:
+  model_name: vit_huge
+  embed_dim: 1280
+  load_checkpoint: true
+  checkpoint_folder: experiment_logs/vith14.224-bs.128-ep.300/
+  read_checkpoint: jepa-ep300.pth.tar
+  use_bfloat16: true
+  num_classes: 182
+
+data:
+  batch_size: 128
+  root_path: ./wilds_data
+  num_workers: 10
+  pin_mem: true
+  crop_size: 224
+  crop_scale: [0.3, 1.0]
+  use_random_resized_crop: false
+  use_horizontal_flip: false
+  use_color_distortion: false
+  color_jitter_strength: 0.0
+  use_gaussian_blur: false
+
+mask:
+  patch_size: 14
+
+optimization:
+  optimizer: adamw # 'adamw' or 'sgd'
+  freeze_weights: false # true for linear probing, false for full fine-tuning
+  epochs: 300 # can be set higher if early_stopping
+  lr: 5.0e-4
+  weight_decay: 1.0e-2
+  use_cosine_schedule: true
+  start_lr: 0.0001
+  final_lr: 1.0e-06
+  warmup: 5
+  ipe_scale: 1.0
+  early_stopping:
+    enabled: true
+    patience: 10
+    min_delta: 1.0e-4
+    min_epochs: 15
+    restore_best_weights: true
+
+validation:
+  eval_every: 1
+
+logging:
+  folder: experiment_logs/supervised-vith14.224-bs.128-ep.300-ft/
+  write_tag: fine_tune

configs/supervised_wilds_vith14_ep300-lp.yaml

@@ -14,6 +14,7 @@ data:
   pin_mem: true
   crop_size: 224
   crop_scale: [0.3, 1.0]
+  use_random_resized_crop: false
   use_horizontal_flip: false
   use_color_distortion: false
   color_jitter_strength: 0.0
@@ -44,5 +45,5 @@ validation:
   eval_every: 1
 
 logging:
-  folder: experiment_logs/supervised-vith14.224-bs.128-ep.300/
+  folder: experiment_logs/supervised-vith14.224-bs.128-ep.300-lp/
   write_tag: linear_probe

src/train.py

@@ -78,6 +78,7 @@ def main(args, resume_preempt=False):
     use_horizontal_flip = args["data"]["use_horizontal_flip"]
     use_color_distortion = args["data"]["use_color_distortion"]
     color_jitter = args["data"]["color_jitter_strength"]
+    use_random_resized_crop = args["data"].get("use_random_resized_crop", True)
     # --
     batch_size = args["data"]["batch_size"]
     pin_mem = args["data"]["pin_mem"]
@@ -181,6 +182,7 @@ def main(args, resume_preempt=False):
         horizontal_flip=use_horizontal_flip,
         color_distortion=use_color_distortion,
         color_jitter=color_jitter,
+        use_random_resized_crop=use_random_resized_crop,
     )
 
     # -- init data-loaders/samplers

src/train_supervised.py

@@ -239,6 +239,7 @@ def main(args, resume_preempt=False):
         color_distortion=d_args["use_color_distortion"],
         color_jitter=d_args["color_jitter_strength"],
         gaussian_blur=d_args["use_gaussian_blur"],
+        use_random_resized_crop=d_args.get("use_random_resized_crop", True),
     )
     val_transform = make_transform_eval(
         crop_size=d_args["crop_size"],

src/transforms.py

@@ -23,6 +23,7 @@ def make_transforms(
     horizontal_flip=False,
     color_distortion=False,
     gaussian_blur=False,
+    use_random_resized_crop=True,
     normalization=((0.485, 0.456, 0.406), (0.229, 0.224, 0.225)),
 ):
     logger.info("making imagenet data transforms")
@@ -36,7 +37,14 @@ def make_transforms(
         return color_distort
 
     transform_list = []
-    transform_list += [transforms.RandomResizedCrop(crop_size, scale=crop_scale)]
+    if use_random_resized_crop:
+        transform_list += [transforms.RandomResizedCrop(crop_size, scale=crop_scale)]
+    else:
+        resize_size = int(crop_size * 256 / 224)
+        transform_list += [
+            transforms.Resize(resize_size),
+            transforms.CenterCrop(crop_size),
+        ]
     if horizontal_flip:
         transform_list += [transforms.RandomHorizontalFlip()]
     if color_distortion: