infer, test=document_fix

paddlenlp/transformers/transformer/modeling.py

@@ -864,6 +864,8 @@ class InferTransformerModel(TransformerModel):
  Specify beam search version. It should be in one
  of [`v1`, `v2`]. If `v2`, need to set `alpha`(default to 0.6) for length
  penalty. Default to `v1`.
+ activation (str, optional):
+ The activation used in FFN. Defaults to "relu".
  kwargs:
  The key word arguments can be `rel_len` and `alpha`:
 
@@ -896,6 +898,7 @@ def __init__(self,
  max_out_len=256,
  output_time_major=False,
  beam_search_version='v1',
+ activation="relu",
  **kwargs):
  args = dict(locals())
  args.pop("self")