Label confidence pair post processor logits as tensor

Author: tomas-gajarsky

facetorch/analyzer/predictor/post.py

@@ -297,20 +297,19 @@ def run(self, preds: torch.Tensor) -> List[Prediction]:
         if isinstance(preds, tuple):
             preds = preds[0]
 
-        # Convert tensor to numpy array once instead of in the loop
-        preds_np = preds.cpu().numpy()
-
-        # Use list comprehension instead of loop for creating pred_list
-        pred_list = [
-            Prediction(
+        pred_list = []
+        for i in range(preds.shape[0]):
+            preds_sample = preds[i]
+            preds_sample_list = preds_sample.cpu().numpy().tolist()
+            other_labels = {
+                label: preds_sample_list[j] + self.offsets[j]
+                for j, label in enumerate(self.labels)
+            }
+            pred = Prediction(
                 label="other",
                 logits=preds_sample,
-                other={
-                    label: preds_np[i, j] + offset
-                    for j, (label, offset) in enumerate(zip(self.labels, self.offsets))
-                },
+                other=other_labels,
             )
-            for i, preds_sample in enumerate(preds_np)
-        ]
+            pred_list.append(pred)
 
         return pred_list