ViLT processor for fine-tuned model

preprocessor_config.json

@@ -1,5 +1,4 @@
 {
-  "do_convert_rgb": true,
   "do_normalize": true,
   "do_pad": true,
   "do_rescale": true,
@@ -9,13 +8,13 @@
     0.5,
     0.5
   ],
-  "image_processor_type": "BlipImageProcessor",
+  "image_processor_type": "ViltImageProcessor",
   "image_std": [
     0.5,
     0.5,
     0.5
   ],
-  "processor_class": "BlipProcessor",
+  "processor_class": "ViltProcessor",
   "resample": 3,
   "rescale_factor": 0.00392156862745098,
   "size": {

tokenizer_config.json

@@ -48,7 +48,7 @@
   "mask_token": "[MASK]",
   "model_max_length": 40,
   "pad_token": "[PAD]",
-  "processor_class": "BlipProcessor",
+  "processor_class": "ViltProcessor",
   "sep_token": "[SEP]",
   "strip_accents": null,
   "tokenize_chinese_chars": true,