Update modules/pmbl.py

modules/pmbl.py

@@ -2,6 +2,7 @@ import sqlite3
 from datetime import datetime
 from llama_cpp import Llama
 from concurrent.futures import ThreadPoolExecutor
+import spaces
 
 class PMBL:
     def __init__(self, model_path):
@@ -101,8 +102,9 @@ class PMBL:
         for chunk in response.result():
             yield chunk
 
+    @spaces.GPU
     def generate_response_task(self, system_prompt, prompt, n_ctx):
-        llm = Llama(model_path=self.model_path, n_ctx=n_ctx, n_threads=8, mlock=True)
+        llm = Llama(model_path=self.model_path, n_ctx=n_ctx, n_threads=8, mlock=True, n_gpu_layers=42)
 
         response = llm(
             system_prompt,
@@ -147,8 +149,9 @@ class PMBL:
 
         conn.close()
 
+    @spaces.GPU
     def generate_topic(self, prompt, response):
-        llm = Llama(model_path=self.model_path, n_ctx=1690, n_threads=8, mlock=True)
+        llm = Llama(model_path=self.model_path, n_ctx=1690, n_threads=8, mlock=True, n_gpu_layers=42)
 
         system_prompt = f"Based on the following interaction between a user and an AI assistant, generate a concise topic for the conversation in 2-4 words:\n\nUser: {prompt}\nAssistant: {response}\n\nTopic:"