fix: tokenizer can be none in preprocessors

processing_aria.py

@@ -18,6 +18,7 @@
 # under the License.
 
 import inspect
+import logging
 import re
 from typing import List, Optional, Union
 
@@ -34,6 +35,8 @@ from transformers.tokenization_utils import (
 
 from .vision_processor import AriaVisionProcessor
 
+logger = logging.getLogger(__name__)
+
 
 class AriaProcessor(ProcessorMixin):
     """
@@ -73,7 +76,7 @@ class AriaProcessor(ProcessorMixin):
         else:
             self.tokenizer = tokenizer
 
-        if self.tokenizer.pad_token is None:
+        if self.tokenizer is not None and self.tokenizer.pad_token is None:
             self.tokenizer.pad_token = self.tokenizer.unk_token
 
         self.image_token = image_token
@@ -229,6 +232,7 @@ class AriaProcessor(ProcessorMixin):
             **cls._extract_kwargs(AriaVisionProcessor.from_pretrained, **kwargs),
         )
         if "use_fast" in kwargs:
+            logger.warning("use_fast is not supported for AriaProcessor. Ignoring...")
             kwargs.pop("use_fast")
         try:
             tokenizer = AutoTokenizer.from_pretrained(