fix(configuration_llama): add `keys_to_ignore_at_inference` to `LlamaConfig` (#23891)

src/transformers/models/llama/configuration_llama.py

@@ -79,6 +79,7 @@ class LlamaConfig(PretrainedConfig):
     >>> configuration = model.config
     ```"""
     model_type = "llama"
+    keys_to_ignore_at_inference = ["past_key_values"]
 
     def __init__(
         self,