Update app.py

app.py

@@ -8,7 +8,7 @@ import time
 import sys
 
 os.system("pip install --upgrade pip")
-os.system('''CMAKE_ARGS="-DLLAMA_AVX512=ON -DLLAMA_AVX512_VBMI=ON -DLLAMA_AVX512_VNNI=ON -DLLAMA_FP16_VA=ON" pip install llama-cpp-python==0.2.13''')
+os.system('''CMAKE_ARGS="-DLLAMA_AVX512=ON -DLLAMA_AVX512_VBMI=ON -DLLAMA_AVX512_VNNI=ON -DLLAMA_FP16_VA=ON" pip install llama-cpp-python''')
 
 from huggingface_hub import snapshot_download
 from llama_cpp import Llama
@@ -20,7 +20,7 @@ You are talking to a human User. If the question is meaningless, please explain
 You are based on SEA model, trained by "SSFW NLPark" team, not related to GPT, LLaMA, Meta, Mistral or OpenAI.
 Let's work this out in a step by step way to be sure we have the right answer.\n\n'''
 SYSTEM_TOKEN = 1587
-USER_TOKEN = 2188
+USER_TOKEN = 4096
 BOT_TOKEN = 12435
 LINEBREAK_TOKEN = 13
 
@@ -56,7 +56,7 @@ model = Llama(
     n_parts=1,
 )
 
-max_new_tokens = 1500
+max_new_tokens = 4096
 
 def user(message, history):
     new_history = history + [[message, None]]