add use_normal_tokenizers option

modules/app.py

@@ -1,8 +1,12 @@
 import sys
 from typing import List
 import traceback
+import os
 # needs to be imported *before* transformers
-import tokenizers_patch
+if os.path.exists('use_normal_tokenizers'):
+    import tokenizers
+else:
+    import tokenizers_patch
 from transformers import AutoModelForCausalLM, AutoTokenizer
 import json
 
@@ -20,7 +24,7 @@ import json
 PORT = 7860
 VERBOSE = False
 
-BIG_MODEL = True
+BIG_MODEL = False
 
 if BIG_MODEL:
     CUDA = True