diff --git a/unsloth/models/llama.py b/unsloth/models/llama.py index ec6706e515..511ae5c681 100644 --- a/unsloth/models/llama.py +++ b/unsloth/models/llama.py @@ -724,7 +724,8 @@ def LlamaModel_fast_forward( past_key_values_length, sliding_window = getattr(self.config, "sliding_window", None), ) - attention_mask = attention_mask.to(torch.bool) + if attention_mask is not None: + attention_mask = attention_mask.to(torch.bool) pass hidden_states = inputs_embeds