Update app.py

app.py

@@ -47,7 +47,7 @@ class HfModelWrapper:
         input_text = self.build_prompt(user_input, history)
         input_ids = self.tokenizer.encode(input_text, return_tensors="pt").to("cuda")
 
-        streamer = TextIteratorStreamer(self.tokenizer, skip_prompt=True, skip_special_tokens=True)
+        streamer = TextIteratorStreamer(self.tokenizer, skip_prompt=True)#skip_special_tokens=True)
 
         gen_kwargs = {
             "inputs": input_ids,
@@ -57,7 +57,7 @@ class HfModelWrapper:
             "temperature": 0.1,
             "top_p": 0.9,
             #"repetition_penalty": 1.1,
-            "min_length": 5,
+            "min_length": 10,
         }
 
         thread = Thread(target=self.model.generate, kwargs=gen_kwargs)