answer speed optimizations

app.py

@@ -7,12 +7,12 @@ from huggingface_hub import hf_hub_download  # load from huggingfaces
 CONST_REPO_ID = "TheBloke/Llama-2-7B-Chat-GGML"
 CONST_FILENAME = "llama-2-7b-chat.ggmlv3.q4_1.bin"
 
-N_CTX = 2048
+N_CTX = 1024
 
 llm = Llama(model_path=hf_hub_download(
     repo_id=CONST_REPO_ID,
     filename=CONST_FILENAME),
-    n_ctx=2048
+    n_ctx=N_CTX
 )
 history = N_CTX
 
@@ -29,7 +29,7 @@ def generate_text(input_text, history):
     else:
         input_text_with_history = f"{history[-1][1]}" + "\n"
         input_text_with_history += f"USER: {input_text}" + "\n" + " ASSISTANT:"
-    output = llm(input_text_with_history, max_tokens=4096, stop=[
+    output = llm(input_text_with_history, max_tokens=1024, stop=[
         "<|prompter|>", "<|endoftext|>", "<|endoftext|> \n",
         "ASSISTANT:", "USER:", "SYSTEM:"], stream=True
     )