from transformers import BartForConditionalGeneration, GenerationConfig
import torch

model_id = "hf-internal-testing/tiny-random-bart"
model = BartForConditionalGeneration.from_pretrained(model_id)
gen_config = GenerationConfig(
    num_beams=4,
    do_sample=False,
    max_new_tokens=10,
    min_new_tokens=10,
    num_beam_groups=2,
    diversity_penalty=0.0001,
    use_cache=True,
)

input_ids = torch.tensor([[model.config.bos_token_id]])
out = model.generate(input_ids, generation_config=gen_config, trust_remote_code=True)