diff --git a/vllm/config.py b/vllm/config.py index 4655430dba511..fb322eddc48a1 100644 --- a/vllm/config.py +++ b/vllm/config.py @@ -1820,7 +1820,7 @@ def _get_and_verify_max_len( "original_max_position_embeddings"] derived_max_model_len *= scaling_factor - if bert_config and "max_seq_lenght" in bert_config: + if bert_config and "max_seq_length" in bert_config: derived_max_model_len = bert_config["max_seq_length"] # If the user specified a max length, make sure it is smaller than the