ruffedisorted

Author: Johanmkr

.gitignore

@@ -15,6 +15,9 @@ env2/*
 ruffian.sh 
 localtest.sh
 
+# Johanthings
+formatting.x 
+
 # Byte-compiled / optimized / DLL files
 __pycache__/
 *.py[cod]

CollaborativeCoding/dataloaders/mnist_4_9.py

@@ -20,7 +20,14 @@ class MNISTDataset4_9(Dataset):
         Whether to train the model or not, by default False
     """
 
-    def __init__(self, data_path: Path, sample_ids: np.ndarray, train: bool = False, transform = None, nr_channels: int = 1):
+    def __init__(
+        self,
+        data_path: Path,
+        sample_ids: np.ndarray,
+        train: bool = False,
+        transform=None,
+        nr_channels: int = 1,
+    ):
         super.__init__()
         self.data_path = data_path
         self.mnist_path = self.data_path / "MNIST"
@@ -51,7 +58,7 @@ def __getitem__(self, idx):
             )
 
         image = np.expand_dims(image, axis=0)  # Channel
-        
+
         if self.transform:
             image = self.transform(image)
 

CollaborativeCoding/load_data.py

@@ -68,7 +68,7 @@ def load_data(dataset: str, *args, **kwargs) -> tuple:
         case "mnist_4-9":
             dataset = MNISTDataset4_9
             train_labels, test_labels = downloader.mnist(data_dir=data_dir)
-            labels = np.arange(4,10)
+            labels = np.arange(4, 10)
         case _:
             raise NotImplementedError(f"Dataset: {dataset} not implemented.")
 

CollaborativeCoding/metrics/precision.py

@@ -1,6 +1,6 @@
+import numpy as np
 import torch
 import torch.nn as nn
-import numpy as np
 
 
 class Precision(nn.Module):
@@ -42,7 +42,6 @@ def forward(self, y_true: torch.tensor, logits: torch.tensor) -> torch.tensor:
         # Append to the class-global values
         self.y_true.append(y_true)
         self.y_pred.append(y_pred)
-        
 
     def _micro_avg_precision(
         self, y_true: torch.tensor, y_pred: torch.tensor
@@ -99,17 +98,21 @@ def _macro_avg_precision(
         fp = torch.sum(~true_oh.bool() * pred_oh, 0)
 
         return torch.nanmean(tp / (tp + fp))
-    
+
     def __returnmetric__(self):
         if self.y_true == [] and self.y_pred == []:
             return np.nan
         elif self.y_true == [] or self.y_pred == []:
             raise ValueError("y_true or y_pred is empty.")
         self.y_true = torch.cat(self.y_true)
         self.y_pred = torch.cat(self.y_pred)
-        
-        return self._macro_avg_precision(self.y_true, self.y_pred) if self.macro_averaging else self._micro_avg_precision(self.y_true, self.y_pred)    
-    
+
+        return (
+            self._macro_avg_precision(self.y_true, self.y_pred)
+            if self.macro_averaging
+            else self._micro_avg_precision(self.y_true, self.y_pred)
+        )
+
     def __reset__(self):
         """Resets the class-global lists of true and predicted values to empty lists.
 

tests/test_metrics.py

@@ -98,9 +98,13 @@ def test_precision():
     precision_micro = Precision(num_classes=C)
     precision_macro = Precision(num_classes=C, macro_averaging=True)
 
-    # find scores
-    micro_precision_score = precision_micro(y_true, logits)
-    macro_precision_score = precision_macro(y_true, logits)
+    # run metric object
+    precision_micro(y_true, logits)
+    precision_macro(y_true, logits)
+
+    # get metric scores
+    micro_precision_score = precision_micro.__returnmetric__()
+    macro_precision_score = precision_macro.__returnmetric__()
 
     # check output to be tensor
     assert isinstance(micro_precision_score, torch.Tensor), "Tensor output is expected."