update app

app.py

@@ -60,10 +60,12 @@ def generate_stream(prompt, max_new_tokens=128):
     #     if text.strip():
     #         yield text
             
-    streamer = TextIteratorStreamer(tokenizer, skip_special_tokens=True)
+    streamer = TextIteratorStreamer(tokenizer, skip_prompt=True, skip_special_tokens=True)
     generation_kwargs = dict(
         **inputs,
         max_new_tokens=max_new_tokens,
+        eos_token_id=tokenizer.eos_token_id,
+        do_sample=True,
         temperature=0.7,
         streamer=streamer
     )