Update app.py

app.py

@@ -26,8 +26,8 @@ def generate(
     else:
         prompt = f"<|im_start|>user\n{user_input}<|im_end|>\n<|im_start|>assistant\n"
     streamer = TextIteratorStreamer(pipe.tokenizer, timeout=240.0, skip_prompt=True, skip_special_tokens=True)
-    generation_kwargs = dict(text_inputs=prompt, streamer=streamer, max_new_tokens=max_new_tokens, do_sample=False, top_p=top_p, min_p=min_p, top_k=top_k, 
-                              temperature=temperature, num_beams=1, repetition_penalty=1.2, dola_layers="low")
+    generation_kwargs = dict(text_inputs=prompt, streamer=streamer, max_new_tokens=max_new_tokens, do_sample=True, top_p=top_p, min_p=min_p, top_k=top_k, 
+                              temperature=temperature, num_beams=1, repetition_penalty=1.1)
     t = Thread(target=pipe.__call__, kwargs=generation_kwargs)
     t.start()
     outputs = []