stouputils 1.3.0__tar.gz → 1.3.1__tar.gz

{stouputils-1.3.0 → stouputils-1.3.1}/PKG-INFO

@@ -1,6 +1,6 @@
 Metadata-Version: 2.4
 Name: stouputils
-Version: 1.3.0
+Version: 1.3.1
 Summary: Stouputils is a collection of utility modules designed to simplify and enhance the development process. It includes a range of tools for tasks such as execution of doctests, display utilities, decorators, as well as context managers, and many more.
 Project-URL: Homepage, https://github.com/Stoupy51/stouputils
 Project-URL: Issues, https://github.com/Stoupy51/stouputils/issues

{stouputils-1.3.0 → stouputils-1.3.1}/pyproject.toml

@@ -5,7 +5,7 @@ build-backend = "hatchling.build"
 
 [project]
 name = "stouputils"
-version = "1.3.0"
+version = "1.3.1"
 description = "Stouputils is a collection of utility modules designed to simplify and enhance the development process. It includes a range of tools for tasks such as execution of doctests, display utilities, decorators, as well as context managers, and many more."
 readme = "README.md"
 requires-python = ">=3.10"

{stouputils-1.3.0 → stouputils-1.3.1}/stouputils/applications/upscaler/video.py

@@ -17,6 +17,7 @@ The module includes YouTube's recommended bitrate settings for different resolut
 framerates, and HDR/SDR content types, ensuring optimal quality for various outputs.
 
 Example usage:
+
 .. code-block:: python
 
 	# Imports

{stouputils-1.3.0 → stouputils-1.3.1}/stouputils/data_science/dataset/dataset.py

@@ -271,7 +271,7 @@ class Dataset:
 		yield from (self.training_data, self.val_data, self.test_data)
 
 	def get_experiment_name(self, override_name: str = "") -> str:
-		""" Get the experiment name for mlflow, example: "DatasetName_GroupingStrategyName"
+		""" Get the experiment name for mlflow, e.g. "DatasetName_GroupingStrategyName"
 
 		Args:
 			override_name (str): Override the Dataset name

{stouputils-1.3.0 → stouputils-1.3.1}/stouputils/data_science/models/base_keras.py

@@ -35,7 +35,7 @@ import mlflow.keras
 import numpy as np
 import tensorflow as tf
 from keras.backend import clear_session
-from keras.callbacks import Callback, CallbackList, EarlyStopping, History, ModelCheckpoint, ReduceLROnPlateau, TensorBoard
+from keras.callbacks import Callback, CallbackList, EarlyStopping, History, ReduceLROnPlateau, TensorBoard
 from keras.layers import Dense, GlobalAveragePooling2D
 from keras.losses import CategoricalCrossentropy, CategoricalFocalCrossentropy, Loss
 from keras.metrics import AUC, CategoricalAccuracy, F1Score, Metric
@@ -52,7 +52,7 @@ from .. import mlflow_utils
 from ..config.get import DataScienceConfig
 from ..dataset import Dataset, GroupingStrategy
 from ..utils import Utils
-from .keras_utils.callbacks import ColoredProgressBar, LearningRateFinder, ProgressiveUnfreezing, WarmupScheduler
+from .keras_utils.callbacks import ColoredProgressBar, LearningRateFinder, ModelCheckpointV2, ProgressiveUnfreezing, WarmupScheduler
 from .keras_utils.losses import NextGenerationLoss
 from .keras_utils.visualizations import all_visualizations_for_image
 from .model_interface import ModelInterface
@@ -553,8 +553,9 @@ class BaseKeras(ModelInterface):
 
 		# Create the checkpoint callback
 		os.makedirs(os.path.dirname(checkpoint_path), exist_ok=True)
-		model_checkpoint: ModelCheckpoint = ModelCheckpoint(
-			checkpoint_path,
+		model_checkpoint: ModelCheckpointV2 = ModelCheckpointV2(
+			epochs_before_start=self.model_checkpoint_delay,
+			filepath=checkpoint_path,
 			monitor="val_loss",
 			mode="min",
 			save_best_only=True,

{stouputils-1.3.0 → stouputils-1.3.1}/stouputils/data_science/models/keras_utils/callbacks/__init__.py

@@ -6,13 +6,15 @@ Features:
 - Warmup scheduler callback for warmup training
 - Progressive unfreezing callback for unfreezing layers during training (incompatible with model.fit(), need a custom training loop)
 - Tqdm progress bar callback for better training visualization
+- Model checkpoint callback that only starts checkpointing after a given number of epochs
 """
 
 # Imports
 from .colored_progress_bar import ColoredProgressBar
 from .learning_rate_finder import LearningRateFinder
+from .model_checkpoint_v2 import ModelCheckpointV2
 from .progressive_unfreezing import ProgressiveUnfreezing
 from .warmup_scheduler import WarmupScheduler
 
-__all__ = ["ColoredProgressBar", "LearningRateFinder", "ProgressiveUnfreezing", "WarmupScheduler"]
+__all__ = ["ColoredProgressBar", "LearningRateFinder", "ModelCheckpointV2", "ProgressiveUnfreezing", "WarmupScheduler"]
 

stouputils-1.3.1/stouputils/data_science/models/keras_utils/callbacks/model_checkpoint_v2.py

@@ -0,0 +1,31 @@
+
+# pyright: reportMissingTypeStubs=false
+# pyright: reportUnknownMemberType=false
+
+# Imports
+from typing import Any
+
+from keras.callbacks import ModelCheckpoint
+
+
+class ModelCheckpointV2(ModelCheckpoint):
+	""" Model checkpoint callback but only starts after a given number of epochs.
+
+	Args:
+		epochs_before_start (int): Number of epochs before starting the checkpointing
+	"""
+
+	def __init__(self, epochs_before_start: int = 3, *args: Any, **kwargs: Any) -> None:
+		super().__init__(*args, **kwargs)
+		self.epochs_before_start = epochs_before_start
+		self.current_epoch = 0
+
+	def on_batch_end(self, batch: int, logs: dict[str, Any] | None = None) -> None:
+		if self.current_epoch >= self.epochs_before_start:
+			super().on_batch_end(batch, logs)
+
+	def on_epoch_end(self, epoch: int, logs: dict[str, Any] | None = None) -> None:
+		self.current_epoch = epoch
+		if epoch >= self.epochs_before_start:
+			super().on_epoch_end(epoch, logs)
+

{stouputils-1.3.0 → stouputils-1.3.1}/stouputils/data_science/models/model_interface.py

@@ -26,8 +26,8 @@ from mlflow.entities import Run
 from numpy.typing import NDArray
 from sklearn.utils import class_weight
 
-from ...decorators import handle_error, measure_time, LogLevels
-from ...print import progress, debug, info, warning
+from ...decorators import handle_error, measure_time
+from ...print import progress, debug, info
 from ...ctx import Muffle, MeasureTime
 from ...io import clean_path
 
@@ -109,7 +109,7 @@ class ModelInterface(AbstractModel):
 		self.epochs: int = 50
 		""" Attribute storing the number of epochs for training. """
 		self.class_weight: dict[int, float] | None = None
-		""" Attribute storing the class weights for training, example: {0: 0.34, 1: 0.66}. """
+		""" Attribute storing the class weights for training, e.g. {0: 0.34, 1: 0.66}. """
 
 		# Fine-tuning parameters
 		self.unfreeze_percentage: float = 100
@@ -126,6 +126,8 @@ class ModelInterface(AbstractModel):
 		# Callback parameters
 		self.early_stop_patience: int = 15
 		""" Attribute storing the patience for early stopping. """
+		self.model_checkpoint_delay: int = 0
+		""" Attribute storing the number of epochs before starting the checkpointing. """
 
 		# ReduceLROnPlateau parameters
 		self.learning_rate: float = 1e-4
@@ -162,7 +164,7 @@ class ModelInterface(AbstractModel):
 		""" Attribute storing the number of epochs for the Unfreeze Percentage Finder """
 		self.unfreeze_finder_update_per_epoch: bool = True
 		""" Attribute storing if the Unfreeze Finder should unfreeze every epoch (True) or batch (False). """
-		self.unfreeze_finder_update_interval: int = 5
+		self.unfreeze_finder_update_interval: int = 25
 		""" Attribute storing the number of steps between each unfreeze, bigger value means more stable loss. """
 
 		## Model architecture
@@ -423,10 +425,11 @@ class ModelInterface(AbstractModel):
 
 		# Callback parameters
 		self.early_stop_patience = override.get("early_stop_patience", self.early_stop_patience)
-		self.reduce_lr_patience = override.get("reduce_lr_patience", self.reduce_lr_patience)
+		self.model_checkpoint_delay = override.get("model_checkpoint_delay", self.model_checkpoint_delay)
 
 		# ReduceLROnPlateau parameters
 		self.learning_rate = override.get("learning_rate", self.learning_rate)
+		self.reduce_lr_patience = override.get("reduce_lr_patience", self.reduce_lr_patience)
 		self.min_delta = override.get("min_delta", self.min_delta)
 		self.min_lr = override.get("min_lr", self.min_lr)
 		self.factor = override.get("factor", self.factor)
@@ -506,9 +509,10 @@ class ModelInterface(AbstractModel):
 
 			# Callback parameters
 			"param_early_stop_patience": self.early_stop_patience,
-			"param_reduce_lr_patience": self.reduce_lr_patience,
+			"param_model_checkpoint_delay": self.model_checkpoint_delay,
 
 			# ReduceLROnPlateau parameters
+			"param_reduce_lr_patience": self.reduce_lr_patience,
 			"param_min_delta": self.min_delta,
 			"param_min_lr": self.min_lr,
 			"param_factor": self.factor,
@@ -550,18 +554,10 @@ class ModelInterface(AbstractModel):
 
 		# Verbose info message
 		if verbose > 0:
-			# If there are multiple validation samples or no filepaths, show the number of validation samples
-			if len(dataset.val_data.X) != 1 or not dataset.val_data.filepaths:
-				info(
-					f"({self.model_name}) Training final model on full dataset with "
-					f"{len(dataset.training_data.X)} samples ({len(dataset.val_data.X)} validation)"
-				)
-			# Else, show the filepath of the single validation sample (useful for debugging)
-			else:
-				info(
-					f"({self.model_name}) Training final model on full dataset with "
-					f"{len(dataset.training_data.X)} samples (validation: {dataset.val_data.filepaths[0]})"
-				)
+			info(
+				f"({self.model_name}) Training final model on full dataset with "
+				f"{len(dataset.training_data.X)} samples ({len(dataset.val_data.X)} validation)"
+			)
 
 		# Put the validation data in the test data (since we don't use the test data in the train function)
 		old_test_data: XyTuple = dataset.test_data
@@ -823,7 +819,7 @@ class ModelInterface(AbstractModel):
 
 		# Prepare visualization arguments if needed
 		temp_dir: TemporaryDirectory[str] | None = None
-		if DataScienceConfig.DO_SALIENCY_AND_GRADCAM and dataset.val_data.n_samples == 1:
+		if DataScienceConfig.DO_SALIENCY_AND_GRADCAM and dataset.test_data.n_samples == 1:
 			temp_dir = TemporaryDirectory()
 
 		# Create and run the process