Adds config yaml

conf/config.yaml

@@ -0,0 +1,18 @@
+# data:
+train_csv_path: data/diabetic-retinopathy-dataset/train.csv
+val_csv_path: data/diabetic-retinopathy-dataset/val.csv
+
+# experiment:
+seed: 42
+batch_size: 128
+num_workers: 2
+use_class_weighting: false
+use_weighted_sampler: false  # class weighting and weighted sampler are mutually exclusive
+model_name: "resnet50"
+max_epochs: 50
+image_size: 224
+learning_rate: 3e-4
+use_scheduler: true
+
+logs_dir: logs
+checkpoint_dirpath: artifacts/checkpoints

config.py

@@ -1,29 +0,0 @@
-from dataclasses import dataclass
-
-@dataclass
-class Config:
-    # Data paths
-    train_csv_path: str = "data/train.csv"
-    val_csv_path: str = "data/val.csv"
-    test_csv_path: str = "data/test.csv"  # Added for test data
-    
-    # Model parameters
-    input_size: tuple[int, int] = (224, 224)  # Input image size
-    num_classes: int = 5  # Number of output classes
-    batch_size: int = 32
-    epochs: int = 10
-    learning_rate: float = 0.001
-    model_architecture: str = "PretrainedResNet50"  # Specify the backbone architecture
-    loss_function: str = "cross_entropy"
-    optimizer: str = "Adam"
-    lr_scheduler: str = "StepLR"
-    dropout_rate: float = 0.5
-    weight_decay: float = 0.001
-    early_stopping: bool = True
-    use_gpu: bool = True
-    random_seed: int = 42
-    data_augmentation: bool = True
-    
-    # Model paths
-    model_save_path: str = "models/model.pth"
-    logs_path: str = "logs/"