Update utils.py

Use a wordaround method to solve the space missing issue for llama-based tokenizer

utils.py

@@ -16,7 +16,22 @@ def get_res(model_name, input_sentence, single_print=True):
     out = tokenizer.encode(input_sentence, add_special_tokens=False)
     token_num = len(out)
 
-    w = [ f'<span style="font-size:1.25em;background-color:{next(color_iterator)}">{tokenizer.decode(x)}</span>' for x in out ]
+    work_around = False
+    if work_around:
+        w = [] 
+        pre = ""
+        for i in range(len(out)): 
+            res = tokenizer.decode(out[:i+1])  
+            if w == []:
+                w.append(res)
+            else:
+                pre_len = len(pre) #0
+                w.append(res[pre_len:])  
+            pre = res
+                
+        w = [ f'<span style="font-size:1.25em;background-color:{next(color_iterator)}">{x}</span>' for x in out ]
+    else:
+        w = [ f'<span style="font-size:1.25em;background-color:{next(color_iterator)}">{tokenizer.decode(x)}</span>' for x in out ]
     res = ''.join(w)
     if single_print:
         print(res + str(token_num))