diff --git a/python/llm/src/bigdl/llm/transformers/speculative.py b/python/llm/src/bigdl/llm/transformers/speculative.py index 9942398d..e3c8cfa2 100644 --- a/python/llm/src/bigdl/llm/transformers/speculative.py +++ b/python/llm/src/bigdl/llm/transformers/speculative.py @@ -57,7 +57,7 @@ def generate( new_speculative_kwargs = {} for var in ['max_new_tokens', 'max_step_draft', 'th_stop_draft', 'do_sample', 'top_k', 'top_p', 'temperature', 'hf_adjust', - 'auto_th_stop_draft', 'auto_parameters']: + 'auto_th_stop_draft', 'auto_parameters', 'repetition_penalty']: value = kwargs.pop(var, None) if value is not None: new_speculative_kwargs[var] = value