Update app.py

app.py

@@ -4,18 +4,25 @@ os.system('CMAKE_ARGS="-DLLAMA_OPENBLAS=on" FORCE_CMAKE=1 pip install llama-cpp-
 import wget
 from llama_cpp import Llama
 import random
-url = 'https://huggingface.co/TheBloke/WizardLM-7B-uncensored-GGML/resolve/main/WizardLM-7B-uncensored.ggmlv3.q2_K.bin'
+
+def get_num_cores():
+    """Get the number of CPU cores."""
+    return os.cpu_count()
+
+if __name__ == "__main__":
+    num_cores = get_num_cores()
+url = 'https://huggingface.co/TheBloke/Llama-2-7B-Chat-GGML/resolve/main/llama-2-7b-chat.ggmlv3.q2_K.bin'
 filename = wget.download(url)
-llm2 = Llama(model_path=filename, seed=random.randint(1, 2**31))
+llm2 = Llama(model_path=filename, seed=random.randint(1, 2**31), lora_path="ggml-adapter-model.bin", n_threads=num_cores)
 filename = wget.download(url)
 theme = gr.themes.Soft(
     primary_hue=gr.themes.Color("#ededed", "#fee2e2", "#fecaca", "#fca5a5", "#f87171", "#ef4444", "#dc2626", "#b91c1c", "#991b1b", "#7f1d1d", "#6c1e1e"),
     neutral_hue="red",
 )
-title = """<h1 align="center">Chat with awesome WizardLM 7b model!</h1><br>"""
+title = """<h1 align="center">Chat with awesome LLAMA 2 CHAT model!</h1><br>"""
 with gr.Blocks(theme=theme) as demo:
     gr.HTML(title)
-    gr.HTML("This model is awesome for its size! It is only 20th the size of Chatgpt but is around 90% as good as Chatgpt. However, please don't rely on WizardLM to provide 100% true information as it might be wrong sometimes. ")
+    gr.HTML("This model is awesome for its size! It is only 20th the size of Chatgpt but is still decent for chatting. However like all models, LLAMA-2-CHAT can hallucinate and provide incorrect information.")
     chatbot = gr.Chatbot()
     msg = gr.Textbox()
     clear = gr.ClearButton([msg, chatbot])
@@ -30,9 +37,9 @@ with gr.Blocks(theme=theme) as demo:
         #token1 = llm.tokenize(b"### Instruction: ")
         #token2 = llm.tokenize(instruction.encode())
         #token3 = llm2.tokenize(b"USER: ")
-        tokens3 = llm2.tokenize(user_message.encode())
-        token4 = llm2.tokenize(b"\n\n### Response:")
-        tokens = tokens3 + token4
+        #tokens3 = llm2.tokenize(user_message.encode())
+        #token4 = llm2.tokenize(b"\n\n### Response:")
+        tokens = llm2.tokenize(user_message.encode())
         history[-1][1] = ""
         count = 0
         output = ""