diff --git a/unsloth/models/llama.py b/unsloth/models/llama.py index 0b567b023a..8436ab18e8 100644 --- a/unsloth/models/llama.py +++ b/unsloth/models/llama.py @@ -707,9 +707,9 @@ def LlamaModel_fast_forward( # Ignore attention_mask if attention_mask is None: padding_mask = None - elif self.training: - attention_mask = None - padding_mask = None + # elif self.training: + # attention_mask = None + # padding_mask = None else: # if 0 in attention_mask: # padding_mask = attention_mask