Issues attacking se2seq model (seq2sick)

[SachJbp]

Seems like raw outputs of the model are being passed to NonOverlappingOutput implemented methods _is_goal_complete which subsequently calls word_difference_score with those tensors leading to split() being called on tensors giving TypeError: split() missing 1 required positional argument: 'split_size'. Let me know if there is something wrong while constructing the model_wrapper or if BART can be attacked using TextAttack?

import transformers
from transformers import BartForConditionalGeneration, BartTokenizer
import textattack
from textattack.attack_recipes.seq2sick_cheng_2018_blackbox import Seq2SickCheng2018BlackBox

# Define the input text and target summary
input_text = "ministers from the european union and its mediterranean neighbors gathered here under heavy security on monday for an unprecedented conference"
original_summary = "eu mediterranean nations meet for first-ever conference on cooperation."

# Load the Hugging Face BART model
# Load model, tokenizer, and model_wrapper
model = transformers.BartForConditionalGeneration.from_pretrained("facebook/bart-large-cnn")
tokenizer = transformers.BartTokenizer.from_pretrained("facebook/bart-large-cnn")
model_wrapper = textattack.models.wrappers.HuggingFaceModelWrapper(model, tokenizer)

# Specify the attack recipe and constraints
attack_recipe = Seq2SickCheng2018BlackBox.build(model_wrapper)

# Run the attack
result = attack_recipe.attack(input_text, original_summary)

# Print the result
print(result.__str__(color_method="ansi"))