app

app.py

@@ -43,6 +43,7 @@ def mlm(image, text):
                 predicted_tokens.append(tokenizer.convert_ids_to_tokens([idx])[0])
             # for i in range(1):
             #     res[text][k].append(predicted_tokens)
+            print(predicted_tokens)
             res[k] = predicted_tokens[0]
         color, shape, loc = res['color'], res['shape'], res['location']
         ans = f'{color} color, {shape} shape, cat at {loc}'