generation_config.json

{
  "decoder_start_token_id": 2,
  "do_sample": true,
  "max_length": 200,
  "max_new_tokens": 200,
  "num_beams": 5,
  "transformers_version": "4.46.3"
}