chore: fill missing docstrings (#162)

finetuner/labeler/__init__.py

@@ -20,6 +20,21 @@ def fit(
     loss: str = 'CosineSiameseLoss',
     **kwargs,
 ) -> None:
+    """Fit the model in an interactive UI.
+
+    :param embed_model: The embedding model to fine-tune
+    :param train_data: Data on which to train the model
+    :param clear_labels_on_start: If set True, will remove all labeled data.
+    :param port_expose: The port to expose.
+    :param runtime_backend: The parallel backend of the runtime inside the Pea, either ``thread`` or ``process``.
+    :param loss: Which loss to use in training. Supported
+        losses are:
+        - ``CosineSiameseLoss`` for Siamese network with cosine distance
+        - ``EuclideanSiameseLoss`` for Siamese network with eculidean distance
+        - ``CosineTripletLoss`` for Triplet network with cosine distance
+        - ``EuclideanTripletLoss`` for Triplet network with eculidean distance
+    :param kwargs: Additional keyword arguments.
+    """
     dam_path = tempfile.mkdtemp()
 
     class MyExecutor(FTExecutor):

finetuner/tailor/__init__.py

@@ -32,6 +32,17 @@ def to_embedding_model(
     input_dtype: str = 'float32',
     **kwargs
 ) -> AnyDNN:
+    """Convert a general model from :py:attr:`.model` to an embedding model.
+
+    :param model: The DNN model to be converted.
+    :param layer_name: the name of the layer that is used for output embeddings. All layers *after* that layer
+        will be removed. When set to ``None``, then the last layer listed in :py:attr:`.embedding_layers` will be used.
+        To see all available names you can check ``name`` field of :py:attr:`.embedding_layers`.
+    :param output_dim: the dimensionality of the embedding output.
+    :param freeze: if set, then freeze all weights of the original model.
+    :param input_size: The input size of the DNN model.
+    :param input_dtype: The input data type of the DNN model.
+    """
     ft = _get_tailor_class(model)
 
     return ft(model, input_size, input_dtype).to_embedding_model(
@@ -44,6 +55,12 @@ def display(
     input_size: Optional[Tuple[int, ...]] = None,
     input_dtype: str = 'float32',
 ) -> AnyDNN:
+    """Display the model architecture from :py:attr:`.summary` in a table.
+
+    :param model: The DNN model to display.
+    :param input_size: The input size of the DNN model.
+    :param input_dtype: The input data type of the DNN model.
+    """
     ft = _get_tailor_class(model)
 
     return ft(model, input_size, input_dtype).display()

finetuner/tailor/keras/__init__.py

@@ -11,6 +11,12 @@ class KerasTailor(BaseTailor):
     """Tailor class for Keras DNN models."""
 
     def summary(self, skip_identity_layer: bool = False) -> LayerInfoType:
+        """Interpret the DNN model and produce model information.
+
+        :param skip_identity_layer: If skip identity layer.
+        :return: The model information stored as dict.
+        """
+
         def _get_output_shape(layer):
             try:
                 return layer.output_shape
@@ -70,6 +76,16 @@ def to_embedding_model(
         freeze: bool = False,
     ) -> AnyDNN:
 
+        """Convert a general model from :py:attr:`.model` to an embedding model.
+
+        :param layer_name: the name of the layer that is used for output embeddings. All layers *after* that layer
+            will be removed. When set to ``None``, then the last layer listed in :py:attr:`.embedding_layers` will be used.
+            To see all available names you can check ``name`` field of :py:attr:`.embedding_layers`.
+        :param output_dim: the dimensionality of the embedding output.
+        :param freeze: if set, then freeze all weights of the original model.
+        :return: Converted embedding model.
+        """
+
         if layer_name:
             _all_embed_layers = {l['name']: l for l in self.embedding_layers}
             try:

finetuner/tailor/paddle/__init__.py

@@ -19,6 +19,11 @@ class PaddleTailor(BaseTailor):
     """
 
     def summary(self, skip_identity_layer: bool = False) -> LayerInfoType:
+        """Interpret the DNN model and produce model information.
+
+        :param skip_identity_layer: If skip identity layer.
+        :return: The model information stored as dict.
+        """
         if not self._input_size:
             raise ValueError(
                 f'{self.__class__} requires a valid `input_size`, but receiving {self._input_size}'
@@ -134,6 +139,15 @@ def to_embedding_model(
         output_dim: Optional[int] = None,
         freeze: bool = False,
     ) -> AnyDNN:
+        """Convert a general model from :py:attr:`.model` to an embedding model.
+
+        :param layer_name: the name of the layer that is used for output embeddings. All layers *after* that layer
+            will be removed. When set to ``None``, then the last layer listed in :py:attr:`.embedding_layers` will be used.
+            To see all available names you can check ``name`` field of :py:attr:`.embedding_layers`.
+        :param output_dim: the dimensionality of the embedding output.
+        :param freeze: if set, then freeze all weights of the original model.
+        :return: Converted embedding model.
+        """
         model = copy.deepcopy(self._model)
 
         if layer_name:

finetuner/tailor/pytorch/__init__.py

@@ -16,6 +16,11 @@ class PytorchTailor(BaseTailor):
     """Tailor class for PyTorch DNN models"""
 
     def summary(self, skip_identity_layer: bool = False) -> LayerInfoType:
+        """Interpret the DNN model and produce model information.
+
+        :param skip_identity_layer: If skip identity layer.
+        :return: The model information stored as dict.
+        """
         if not self._input_size:
             raise ValueError(
                 f'{self.__class__} requires a valid `input_size`, but receiving {self._input_size}'
@@ -129,6 +134,15 @@ def to_embedding_model(
         output_dim: Optional[int] = None,
         freeze: bool = False,
     ) -> AnyDNN:
+        """Convert a general model from :py:attr:`.model` to an embedding model.
+
+        :param layer_name: the name of the layer that is used for output embeddings. All layers *after* that layer
+            will be removed. When set to ``None``, then the last layer listed in :py:attr:`.embedding_layers` will be used.
+            To see all available names you can check ``name`` field of :py:attr:`.embedding_layers`.
+        :param output_dim: the dimensionality of the embedding output.
+        :param freeze: if set, then freeze all weights of the original model.
+        :return: Converted embedding model.
+        """
 
         model = copy.deepcopy(self._model)
 

finetuner/tuner/__init__.py

@@ -44,6 +44,12 @@ def fit(
     :param eval_data: Data on which to evaluate the model at the end of each epoch
     :param epochs: Number of epochs to train the model
     :param batch_size: The batch size to use for training and evaluation
+    :param loss: Which loss to use in training. Supported
+        losses are:
+        - ``CosineSiameseLoss`` for Siamese network with cosine distance
+        - ``EuclideanSiameseLoss`` for Siamese network with eculidean distance
+        - ``CosineTripletLoss`` for Triplet network with cosine distance
+        - ``EuclideanTripletLoss`` for Triplet network with eculidean distance
     :param learning_rate: Learning rate to use in training
     :param optimizer: Which optimizer to use in training. Supported
         values/optimizers are:

finetuner/tuner/keras/losses.py

@@ -34,7 +34,7 @@ def call(self, inputs, **kwargs):
 
 
 class EuclideanSiameseLoss(BaseLoss, Layer):
-    """Computes the loss for a siamese network using cosine distance.
+    """Computes the loss for a siamese network using eculidean distance.
 
     This loss is also known as contrastive loss.
 

finetuner/tuner/paddle/losses.py

@@ -37,7 +37,7 @@ def forward(
 
 
 class EuclideanSiameseLoss(BaseLoss, nn.Layer):
-    """Computes the loss for a siamese network using cosine distance.
+    """Computes the loss for a siamese network using eculidean distance.
 
     This loss is also known as contrastive loss.
 

finetuner/tuner/pytorch/losses.py

@@ -37,7 +37,7 @@ def forward(
 
 
 class EuclideanSiameseLoss(BaseLoss, nn.Module):
-    """Computes the loss for a siamese network using cosine distance.
+    """Computes the loss for a siamese network using eculidean distance.
 
     This loss is also known as contrastive loss.