Let's call it batch ratio

embedding_converter/README.md

@@ -32,9 +32,9 @@ file_pattern = .datasets/megaface/**/*.jpg
 
 ```
 [training.loader]
-split_ratio = 0.95
 batch_size = 256
 num_workers = 8
+split_ratio = 0.95
 ```
 
 ```

embedding_converter/config.ini

@@ -2,9 +2,9 @@
 file_pattern =
 
 [training.loader]
-split_ratio =
 batch_size =
 num_workers =
+split_ratio =
 
 [training.model]
 source_path =

face_swapper/README.md

@@ -28,7 +28,7 @@ This `config.ini` utilizes the MegaFace dataset to train the Face Swapper model.
 ```
 [training.dataset]
 file_pattern = .datasets/vggface2/**/*.jpg
-equal_probability = 0.2
+batch_ratio = 0.2
 ```
 
 ```

face_swapper/config.ini

@@ -1,6 +1,6 @@
 [training.dataset]
 file_pattern =
-equal_probability =
+batch_ratio =
 
 [training.loader]
 batch_size =

face_swapper/src/dataset.py

@@ -10,15 +10,15 @@ from .types import Batch
 
 
 class DynamicDataset(Dataset[Tensor]):
-	def __init__(self, file_pattern : str, equal_probability : float) -> None:
+	def __init__(self, file_pattern : str, batch_ratio : float) -> None:
 		self.file_paths = glob.glob(file_pattern)
 		self.transforms = self.compose_transforms()
-		self.equal_probability = equal_probability
+		self.batch_ratio = batch_ratio
 
 	def __getitem__(self, index : int) -> Batch:
 		source_image_path = self.file_paths[index]
 
-		if random.random() < self.equal_probability:
+		if random.random() < self.batch_ratio:
 			return self.prepare_equal_batch(source_image_path)
 
 		return self.prepare_different_batch(source_image_path)

face_swapper/src/training.py

@@ -143,10 +143,10 @@ def create_trainer() -> Trainer:
 
 def train() -> None:
 	dataset_file_pattern = CONFIG.get('training.dataset', 'file_pattern')
-	dataset_equal_probability = CONFIG.getfloat('training.dataset', 'equal_probability')
+	dataset_batch_ratio = CONFIG.getfloat('training.dataset', 'batch_ratio')
 	output_resume_path = CONFIG.get('training.output', 'resume_path')
 
-	dataset = DynamicDataset(dataset_file_pattern, dataset_equal_probability)
+	dataset = DynamicDataset(dataset_file_pattern, dataset_batch_ratio)
 	training_loader, validation_loader = create_loaders(dataset)
 	face_swapper_trainer = FaceSwapperTrainer()
 	trainer = create_trainer()