from transformers import AutoTokenizer, AutoModelForCausalLM

model_name = "roneneldan/TinyStories-1M"

def load():
  global model
  global tokenizer
  
  model = AutoModelForCausalLM.from_pretrained(model_name)
  tokenizer = AutoTokenizer.from_pretrained(model_name)

def generate(input_text):
  input_ids = tokenizer.encode(input_text, return_tensors="pt")
  output_ids = model.generate(
    input_ids,
    no_repeat_ngram_size=2,
    max_new_tokens=200,
    eos_token_id=tokenizer.eos_token_id,
    temperature=0.2
  )
  
  return tokenizer.decode(output_ids[0], skip_special_tokens=True)