tested

app.py

@@ -13,9 +13,9 @@ def mlm(image, text):
         # 'location': f'[CLS] The {cls_name} normally appears at or near the [MASK] of a cell. [SEP]',
         # 'color': f'[CLS] When a cell is histologically stained, the {cls_name} are in [MASK] color. [SEP]',
         # 'shape': f'[CLS] Mostly the shape of {cls_name} is [MASK]. [SEP]',
-        'location': f'This {text} is at [MASK] place.',
-        'color': f' This {text} is in [MASK] color.',
-        'shape': f' This {text} is in [MASK] shape.',
+        'location': f'This {text} is at [MASK] place',
+        'color': f'This {text} is in [MASK] color',
+        'shape': f'This {text} is in [MASK] shape',
         #'def': f'{cls_name} is a  . [SEP]',
     }
     ans = list()
@@ -23,6 +23,7 @@ def mlm(image, text):
     device = 'cpu'
     for k, v in questions_dict.items():
         predicted_tokens = []
+        print(v)
         tokenized_text = tokenizer.tokenize(v)
         indexed_tokens = tokenizer.convert_tokens_to_ids(tokenized_text)
         # Create the segments tensors.