Update llama.py

This commit is contained in:
Daniel Han-Chen 2024-02-04 02:35:17 +11:00
commit 607dfa1d0e

View file

@ -664,7 +664,7 @@ def LlamaForCausalLM_fast_forward(
*args, **kwargs,
) -> Union[Tuple, CausalLMOutputWithPast]:
if causal_mask is None and past_key_value is not None:
if causal_mask is None and past_key_values is not None:
causal_mask = xformers.attn_bias.LowerTriangularMask()
output_attentions = output_attentions if output_attentions is not None else self.config.output_attentions