Merge pull request #141 from justusschock/code-cleanup

Remove mutable default arguments

delira/data_loading/data_manager.py

@@ -1,19 +1,16 @@
 import inspect
 import logging
 
-import numpy as np
-import typing
-import inspect
 from batchgenerators.dataloading import MultiThreadedAugmenter, \
     SingleThreadedAugmenter, SlimDataLoaderBase
 from batchgenerators.transforms import AbstractTransform
 
+from delira import get_current_debug_mode
 from .data_loader import BaseDataLoader
 from .dataset import AbstractDataset, BaseCacheDataset, BaseLazyDataset
 from .load_utils import default_load_fn_2d
 from .sampler import SequentialSampler, AbstractSampler
 from ..utils.decorators import make_deprecated
-from delira import get_current_debug_mode
 
 logger = logging.getLogger(__name__)
 
@@ -245,7 +242,7 @@ class BaseDataManager(object):
 
     def __init__(self, data, batch_size, n_process_augmentation,
                  transforms, sampler_cls=SequentialSampler,
-                 sampler_kwargs={},
+                 sampler_kwargs=None,
                  data_loader_cls=None, dataset_cls=None,
                  load_fn=default_load_fn_2d, from_disc=True, **kwargs):
         """
@@ -292,6 +289,8 @@ class defining the sampling strategy
         """
 
         # Instantiate Hidden variables for property access
+        if sampler_kwargs is None:
+            sampler_kwargs = {}
         self._batch_size = None
         self._n_process_augmentation = None
         self._transforms = None

delira/data_loading/dataset.py

@@ -583,8 +583,13 @@ class Nii3DLazyDataset(BaseLazyDataset):
         """
 
     @make_deprecated('LoadSample')
-    def __init__(self, data_path, load_fn, img_extensions, gt_extensions,
-                 img_files, label_file, **load_kwargs):
+    def __init__(
+            self,
+            data_path,
+            load_fn,
+            img_files,
+            label_file,
+            **load_kwargs):
         """
          Parameters
         ----------
@@ -639,7 +644,7 @@ class Nii3DCacheDatset(BaseCacheDataset):
      """
 
     @make_deprecated('LoadSample')
-    def __init__(self, data_path, load_fn, img_extensions, gt_extensions,
+    def __init__(self, data_path, load_fn,
                  img_files, label_file, **load_kwargs):
         """
          Parameters

delira/data_loading/load_utils.py

@@ -78,7 +78,11 @@ def is_valid_image_file(fname, img_extensions, gt_extensions):
     ----------
     fname : str
         filename of image path
-     Returns
+    img_extensions : list
+        list of valid image file extensions
+    gt_extensions : list
+        list of valid gt file extensions
+    Returns
     -------
     bool
         is valid data sample
@@ -148,7 +152,7 @@ class LoadSample:
     def __init__(self,
                  sample_ext: dict,
                  sample_fn: collections.abc.Callable,
-                 dtype={}, normalize=(), norm_fn=norm_range('-1,1'),
+                 dtype=None, normalize=(), norm_fn=norm_range('-1,1'),
                  **kwargs):
         """
 
@@ -185,6 +189,8 @@ def __init__(self,
         >>>                                       'seg': 'uint8'},
         >>>                      normalize=('data',))
         """
+        if dtype is None:
+            dtype = {}
         self._sample_ext = sample_ext
         self._sample_fn = sample_fn
         self._dtype = dtype
@@ -235,7 +241,7 @@ def __init__(self,
                  sample_fn: collections.abc.Callable,
                  label_ext: collections.abc.Iterable,
                  label_fn: collections.abc.Callable,
-                 sample_kwargs={}, **kwargs):
+                 sample_kwargs=None, **kwargs):
         """
         Load sample and label from folder
 
@@ -264,6 +270,9 @@ def __init__(self,
         --------
         :class: `LoadSample`
         """
+        if sample_kwargs is None:
+            sample_kwargs = {}
+
         super().__init__(sample_ext, sample_fn, **sample_kwargs)
         self._label_ext = label_ext
         self._label_fn = label_fn
@@ -282,7 +291,7 @@ def __call__(self, path):
         dict
             dict with data and label
         """
-        sample_dict = super(LoadSampleLabel, self).__call__(path)
+        sample_dict = super().__call__(path)
         label_dict = self._label_fn(os.path.join(path, self._label_ext),
                                     **self._label_kwargs)
         sample_dict.update(label_dict)

delira/io/torch.py

@@ -10,7 +10,7 @@
     import torch
     from ..models import AbstractPyTorchNetwork
 
-    def save_checkpoint(file: str, model=None, optimizers={},
+    def save_checkpoint(file: str, model=None, optimizers=None,
                         epoch=None, **kwargs):
         """
         Save model's parameters
@@ -28,6 +28,8 @@ def save_checkpoint(file: str, model=None, optimizers={},
             current epoch (will also be pickled)
 
         """
+        if optimizers is None:
+            optimizers = {}
         if isinstance(model, torch.nn.DataParallel):
             _model = model.module
         else:

delira/models/abstract_network.py

@@ -53,8 +53,8 @@ def __call__(self, *args, **kwargs):
 
     @staticmethod
     @abc.abstractmethod
-    def closure(model, data_dict: dict, optimizers: dict, losses={},
-                metrics={}, fold=0, **kwargs):
+    def closure(model, data_dict: dict, optimizers: dict, losses=None,
+                metrics=None, fold=0, **kwargs):
         """
         Function which handles prediction from batch, logging, loss calculation
         and optimizer step
@@ -90,6 +90,10 @@ def closure(model, data_dict: dict, optimizers: dict, losses={},
             If not overwritten by subclass
 
         """
+        if losses is None:
+            losses = {}
+        if metrics is None:
+            metrics = {}
         raise NotImplementedError()
 
     @staticmethod

delira/models/classification/classification_network.py

@@ -67,7 +67,7 @@ def forward(self, input_batch: torch.Tensor):
 
         @staticmethod
         def closure(model: AbstractPyTorchNetwork, data_dict: dict,
-                    optimizers: dict, losses={}, metrics={},
+                    optimizers: dict, losses=None, metrics=None,
                     fold=0, **kwargs):
             """
             closure method to do a single backpropagation step
@@ -108,6 +108,10 @@ def closure(model: AbstractPyTorchNetwork, data_dict: dict,
 
             """
 
+            if losses is None:
+                losses = {}
+            if metrics is None:
+                metrics = {}
             assert (optimizers and losses) or not optimizers, \
                 "Criterion dict cannot be emtpy, if optimizers are passed"
 

delira/models/classification/classification_network_tf.py

@@ -131,7 +131,7 @@ def _build_model(n_outputs: int, **kwargs):
 
     @staticmethod
     def closure(model: typing.Type[AbstractTfNetwork], data_dict: dict,
-                metrics={}, fold=0, **kwargs):
+                metrics=None, fold=0, **kwargs):
         """
         closure method to do a single prediction.
         This is followed by backpropagation or not based state of
@@ -163,9 +163,10 @@ def closure(model: typing.Type[AbstractTfNetwork], data_dict: dict,
 
         """
 
+        if metrics is None:
+            metrics = {}
         loss_vals = {}
         metric_vals = {}
-        image_names = "input_images"
 
         inputs = data_dict.pop('data')
 

delira/models/gan/generative_adversarial_network.py

@@ -94,7 +94,7 @@ def forward(self, real_image_batch):
 
         @staticmethod
         def closure(model, data_dict: dict,
-                    optimizers: dict, losses={}, metrics={},
+                    optimizers: dict, losses=None, metrics=None,
                     fold=0, **kwargs):
             """
             closure method to do a single backpropagation step
@@ -134,6 +134,10 @@ def closure(model, data_dict: dict,
 
             """
 
+            if losses is None:
+                losses = {}
+            if metrics is None:
+                metrics = {}
             loss_vals = {}
             metric_vals = {}
             total_loss_discr_real = 0

delira/models/segmentation/unet.py

@@ -7,7 +7,6 @@
     import torch
     import torch.nn.functional as F
     from torch.nn import init
-    import logging
     from ..abstract_network import AbstractPyTorchNetwork
 
     class UNet2dPyTorch(AbstractPyTorchNetwork):
@@ -175,8 +174,8 @@ def forward(self, x):
             return {"pred": x}
 
         @staticmethod
-        def closure(model, data_dict: dict, optimizers: dict, losses={},
-                    metrics={}, fold=0, **kwargs):
+        def closure(model, data_dict: dict, optimizers: dict, losses=None,
+                    metrics=None, fold=0, **kwargs):
             """
             closure method to do a single backpropagation step
 
@@ -216,6 +215,10 @@ def closure(model, data_dict: dict, optimizers: dict, losses={},
 
             """
 
+            if losses is None:
+                losses = {}
+            if metrics is None:
+                metrics = {}
             assert (optimizers and losses) or not optimizers, \
                 "Loss dict cannot be emtpy, if optimizers are passed"
 
@@ -618,8 +621,8 @@ def forward(self, x):
             return {"pred": x}
 
         @staticmethod
-        def closure(model, data_dict: dict, optimizers: dict, losses={},
-                    metrics={}, fold=0, **kwargs):
+        def closure(model, data_dict: dict, optimizers: dict, losses=None,
+                    metrics=None, fold=0, **kwargs):
             """
             closure method to do a single backpropagation step
 
@@ -659,6 +662,10 @@ def closure(model, data_dict: dict, optimizers: dict, losses={},
 
             """
 
+            if losses is None:
+                losses = {}
+            if metrics is None:
+                metrics = {}
             assert (optimizers and losses) or not optimizers, \
                 "Loss dict cannot be emtpy, if optimizers are passed"
 

delira/training/base_trainer.py

@@ -1,18 +1,16 @@
-from abc import abstractmethod
 import logging
+import os
 import pickle
 import typing
 
-from ..data_loading.data_manager import Augmenter
-
-from .predictor import Predictor
-from .callbacks import AbstractCallback
-from ..models import AbstractNetwork
-
 import numpy as np
-import os
 from tqdm import tqdm
+
 from delira.logging import TrixiHandler
+from .callbacks import AbstractCallback
+from .predictor import Predictor
+from ..data_loading.data_manager import Augmenter
+from ..models import AbstractNetwork
 
 logger = logging.getLogger(__name__)
 
@@ -614,7 +612,7 @@ def _update_state(self, new_state):
 
         """
         for key, val in new_state.items():
-            if (key.startswith("__") and key.endswith("__")):
+            if key.startswith("__") and key.endswith("__"):
                 continue
 
             try:
@@ -729,7 +727,7 @@ def _reinitialize_logging(self, logging_type, logging_kwargs: dict):
                             handlers=new_handlers)
 
     @staticmethod
-    def _search_for_prev_state(path, extensions=[]):
+    def _search_for_prev_state(path, extensions=None):
         """
         Helper function to search in a given path for previous epoch states
         (indicated by extensions)
@@ -752,6 +750,8 @@ def _search_for_prev_state(path, extensions=[]):
             the latest epoch (1 if no checkpoint was found)
 
         """
+        if extensions is None:
+            extensions = []
         files = []
         for file in os.listdir(path):
             for ext in extensions:

delira/training/parameters.py

@@ -17,10 +17,8 @@ class Parameters(LookupConfig):
 
     """
 
-    def __init__(self, fixed_params={"model": {},
-                                     "training": {}},
-                 variable_params={"model": {},
-                                  "training": {}}):
+    def __init__(self, fixed_params=None,
+                 variable_params=None):
         """
 
         Parameters
@@ -31,6 +29,13 @@ def __init__(self, fixed_params={"model": {},
             variable parameters (can be variated by a hyperparameter search)
         """
 
+        if variable_params is None:
+            variable_params = {"model": {},
+                               "training": {}}
+        if fixed_params is None:
+            fixed_params = {"model": {},
+                            "training": {}}
+
         super().__init__(fixed=fixed_params,
                          variable=variable_params)
 
@@ -238,8 +243,8 @@ def update(self, dict_like, deep=False, ignore=None,
             overwrite a 'regular' value with a dict/Config at lower levels.
             See examples for an illustration of the difference
 
-        Examples:
-        ---------
+        Examples
+        --------
         The following illustrates the update behaviour if
         :obj:allow_dict_overwrite is active. If it isn't, an AttributeError
         would be raised, originating from trying to update "string"::