python run_fine_tuning_peft.py \
      --dataset_name carblacac/twitter-sentiment-analysis \
      --model_name_or_path google/flan-t5-xl \
      --use_lora True \
      --do_train True \
      --do_eval True \
      --do_predict True

  from peft import PeftModel, PeftConfig
  from transformers import AutoModelForSeq2SeqLM, AutoTokenizer

  model_name = "thainq107/flan-t5-xl-twitter-sentiment-analysis-lora"

  config = PeftConfig.from_pretrained(model_name)
  model = AutoModelForSeq2SeqLM.from_pretrained(config.base_model_name_or_path)
  model = PeftModel.from_pretrained(model, model_name)
  tokenizer = AutoTokenizer.from_pretrained(config.base_model_name_or_path)

  inputs = tokenizer("I hate you:", return_tensors="pt")
  outputs = model.generate(**inputs)
  tokenizer.batch_decode(outputs, skip_special_tokens=True)

  python run_fine_tuning_peft.py \
      --dataset_name carblacac/twitter-sentiment-analysis \
      --model_name_or_path google/flan-t5-xl \
      --use_prefix True \
      --do_train True \
      --do_eval True \
      --do_predict True

  python run_fine_tuning_peft.py \
      --dataset_name carblacac/twitter-sentiment-analysis \
      --model_name_or_path google/flan-t5-xl \
      --use_prompt True \
      --do_train True \
      --do_eval True \
      --do_predict True