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 -DLLAMA_WASM_SIMD=ON" pip install llama-cpp-python''')
+os.system('''CMAKE_ARGS="-DLLAMA_AVX512=ON -DLLAMA_AVX512_VBMI=ON -DLLAMA_AVX512_VNNI=ON -DLLAMA_AVX_VNNI=ON -DLLAMA_FP16_VA=ON -DLLAMA_WASM_SIMD=ON" pip install llama-cpp-python''')
 
 from huggingface_hub import snapshot_download
 from llama_cpp import Llama
@@ -46,7 +46,7 @@ def get_system_tokens(model):
 
 
 repo_name = "LoneStriker/openbuddy-gemma-7b-v19.1-4k-GGUF"
-model_name = "openbuddy-gemma-7b-v19.1-4k-Q3_K_L.gguf"
+model_name = "openbuddy-gemma-7b-v19.1-4k-Q5_K_M.gguf"
 
 snapshot_download(repo_id=repo_name, local_dir=".", allow_patterns=model_name)