Support streaming

tokenization_qwen.py

@@ -153,6 +153,10 @@ class QWenTokenizer(PreTrainedTokenizer):
         self.box_end_id = self.special_tokens[self.box_end_tag]
         self.quad_start_id = self.special_tokens[self.quad_start_tag]
         self.quad_end_id = self.special_tokens[self.quad_end_tag]
+        self.image_special_tokens = set([
+            self.ref_start_id, self.ref_end_id, self.box_start_id, self.box_end_id,
+            self.quad_start_id, self.quad_end_id,
+        ])
 
         enc = tiktoken.Encoding(
             "Qwen",
@@ -354,7 +358,11 @@ class QWenTokenizer(PreTrainedTokenizer):
         token_ids = _replace_closed_tag(token_ids, self.img_start_id, self.img_end_id, _decode_imgurl)
 
         if skip_special_tokens:
-            token_ids = [i for i in token_ids if i < self.eod_id]
+            if kwargs.get('keep_image_special', False):
+                token_ids = [i for i in token_ids if i < self.eod_id
+                    or i in self.image_special_tokens]
+            else:
+                token_ids = [i for i in token_ids if i < self.eod_id]
         return self.tokenizer.decode(token_ids, errors=errors or self.errors)
 
     def to_list_format(self, text: str):