fix: only print 'cache saved' in verbose mode (abetlen#1668)

lsorber · abetlen · web-flow · commit 9bab46f64357 · 2024-08-12T14:19:26.000-04:00
Co-authored-by: Andrei &lt;abetlen@gmail.com&gt;
diff --git a/llama_cpp/llama.py b/llama_cpp/llama.py
@@ -1523,7 +1523,8 @@ def logit_bias_processor(
                 if self.verbose:
                     print("Llama._create_completion: cache save", file=sys.stderr)
                 self.cache[prompt_tokens + completion_tokens] = self.save_state()
-                print("Llama._create_completion: cache saved", file=sys.stderr)
+                if self.verbose:
+                    print("Llama._create_completion: cache saved", file=sys.stderr)
             return
 
         if self.cache:
