Small update, seems to be working (outside of one small change) out of the box

app.py

@@ -39,6 +39,10 @@ As a derivate work of [Llama-2-7b-chat](https://huggingface.co/meta-llama/Llama-
 this demo is governed by the original [license](https://huggingface.co/spaces/huggingface-projects/llama-2-7b-chat/blob/main/LICENSE.txt) and [acceptable use policy](https://huggingface.co/spaces/huggingface-projects/llama-2-7b-chat/blob/main/USE_POLICY.md).
 """
 
+SYSTEM_PROMPT = """<s>[INST] <<SYS>>
+
+    <</SYS>>"""
+
 def read_pdf_to_documents(file_path):
     doc = fitz.open(file_path)
     documents = []
@@ -50,16 +54,17 @@ def read_pdf_to_documents(file_path):
 
 # Function to update the global system prompt
 def update_system_prompt(new_prompt):
-    global system_prompt
-    system_prompt = new_prompt
+    global SYSTEM_PROMPT
+    SYSTEM_PROMPT = new_prompt
     query_wrapper_prompt = SimpleInputPrompt("{query_str} [/INST]")
     return "System prompt updated."
 
+@spaces.GPU(duration=240)
 def query_model(question):
     llm = HuggingFaceLLM(
         context_window=4096,
         max_new_tokens=256,
-        system_prompt=system_prompt,
+        system_prompt=SYSTEM_PROMPT,
         query_wrapper_prompt=query_wrapper_prompt,
         model=model,
         tokenizer=tokenizer
@@ -101,21 +106,17 @@ if torch.cuda.is_available():
     model = AutoModelForCausalLM.from_pretrained(model_id, torch_dtype=torch.float16, device_map="auto")
     tokenizer = AutoTokenizer.from_pretrained(model_id)
     tokenizer.use_default_system_prompt = False
-
-    system_prompt = """<s>[INST] <<SYS>>
-
-    <</SYS>>"""
     # Throw together the query wrapper
     query_wrapper_prompt = SimpleInputPrompt("{query_str} [/INST]")
     llm = HuggingFaceLLM(context_window=4096,
                      max_new_tokens=256,
-                     system_prompt=system_prompt,
+                     system_prompt=SYSTEM_PROMPT,
                      query_wrapper_prompt=query_wrapper_prompt,
                      model=model, tokenizer=tokenizer)
     embeddings = LangchainEmbedding(HuggingFaceEmbeddings(model_name="all-MiniLM-L6-v2"))
     service_context = ServiceContext.from_defaults(chunk_size=1024, llm=llm, embed_model=embeddings)
     set_global_service_context(service_context)
-    file_path = Path('/content/Full Pamplet.pdf')#make sure to change this to the document path
+    file_path = Path("files/Full Pamplet.pdf")
     documents = read_pdf_to_documents(file_path)
     index = VectorStoreIndex.from_documents(documents)
     query_engine = index.as_query_engine()
@@ -123,7 +124,7 @@ if torch.cuda.is_available():
 
 update_prompt_interface = gr.Interface(
     fn=update_system_prompt,
-    inputs=gr.Textbox(lines=5, placeholder="Enter the system prompt here...", label="System Prompt", value=system_prompt),
+    inputs=gr.Textbox(lines=5, placeholder="Enter the system prompt here...", label="System Prompt", value=SYSTEM_PROMPT),
     outputs=gr.Textbox(label="Status"),
     title="System Prompt Updater",
     description="Update the system prompt used for context."
@@ -142,7 +143,7 @@ query_interface = gr.Interface(
 combined_interface = gr.TabbedInterface([update_prompt_interface, query_interface], ["Update System Prompt", "Query Assistant"])
 
 # Launch the combined interface
-combined_interface.launch()
+#combined_interface.launch()
 
 """
 @spaces.GPU(duration=240)
@@ -248,4 +249,4 @@ with gr.Blocks(css="style.css") as demo:
     gr.Markdown(LICENSE)
 
 if __name__ == "__main__":
-    demo.queue(max_size=20).launch()
+    demo.queue(max_size=20).launch(debug=True)

flagged/log.csv

@@ -0,0 +1,4 @@
+System Prompt,Status,flag,username,timestamp
+"<s>[INST] <<SYS>>
+
+    <</SYS>>",,,,2024-08-15 07:51:11.445523