Bump Python requirement torchvision to 0.21.0

dan/ocr/evaluate.py

@@ -143,7 +143,12 @@ def eval_nerval(
 
         scores = {
             key: {
-                k: round(value * 100, 2) if k in ["P", "R", "F1"] else value
+                k: round(value * 100, 2)
+                if k in ["P", "R", "F1"] and value is not None
+                # Value can be None when
+                # no entity is predicted for a specific entity type
+                # there is no entity in the labels for a specific entity type
+                else (value or 0)
                 for k, value in values.items()
             }
             for key, values in evaluate(ground_truths, predictions, threshold).items()