mapk bug fixed

Author: bartzbeielstein

pyproject.toml

@@ -7,7 +7,7 @@ build-backend = "setuptools.build_meta"
 
 [project]
 name = "spotPython"
-version = "0.2.43"
+version = "0.2.44"
 authors = [
   { name="T. Bartz-Beielstein", email="tbb@bartzundbartz.de" }
 ]

src/spotPython/light/csvmodel.py

@@ -63,7 +63,7 @@ def training_step(self, batch):
         # self.log("train_mapk", self.train_mapk, on_step=True, on_epoch=False)
         return loss
 
-    def validation_step(self, batch, batch_idx):
+    def validation_step(self, batch, batch_idx, prog_bar=False):
         x, y = batch
         logits = self(x)
         # compute cross entropy loss from logits and y
@@ -73,10 +73,10 @@ def validation_step(self, batch, batch_idx):
         acc = accuracy(preds, y, task="multiclass", num_classes=self._L_out)
         self.valid_mapk(logits, y)
         self.log("valid_mapk", self.valid_mapk, on_step=False, on_epoch=True)
-        self.log("val_loss", loss, prog_bar=True)
-        self.log("val_acc", acc, prog_bar=True)
+        self.log("val_loss", loss, prog_bar=prog_bar)
+        self.log("val_acc", acc, prog_bar=prog_bar)
 
-    def test_step(self, batch, batch_idx):
+    def test_step(self, batch, batch_idx, prog_bar=False):
         x, y = batch
         logits = self(x)
         # compute cross entropy loss from logits and y
@@ -85,8 +85,8 @@ def test_step(self, batch, batch_idx):
         acc = accuracy(preds, y, task="multiclass", num_classes=self._L_out)
         self.test_mapk(logits, y)
         self.log("test_mapk", self.test_mapk, on_step=True, on_epoch=True)
-        self.log("val_loss", loss, prog_bar=True)
-        self.log("val_acc", acc, prog_bar=True)
+        self.log("val_loss", loss, prog_bar=prog_bar)
+        self.log("val_acc", acc, prog_bar=prog_bar)
         return loss, acc
 
     def configure_optimizers(self):

src/spotPython/torch/mapk.py

@@ -64,10 +64,11 @@ def update(self, predicted: torch.Tensor, actual: torch.Tensor):
         # Convert predicted to list of lists of indices sorted by confidence score
         _, predicted = predicted.topk(k=self.k, dim=1)
         predicted = predicted.tolist()
-
+        # Code modified according to: "Inplace update to inference tensor outside InferenceMode
+        # is not allowed. You can make a clone to get a normal tensor before doing inplace update."
         score = np.mean([self.apk(p, a, self.k) for p, a in zip(predicted, actual)])
-        self.total += score
-        self.count += 1
+        self.total = self.total + score
+        self.count = self.count + 1
 
     def compute(self):
         """