jhonparra18/gpt2-med-sft-chat-spanish

  from transformers import AutoModelForCausalLM, AutoTokenizer,BitsAndBytesConfig
  import torch

  sys_prompt="Eres un asistente de IA, debes responder amablemente a las preguntas que haga el usuario"
  messages=[{'role':'system','content':sys_prompt},{'role':'user','content':'hola quién eres?'}]
  
  base_model="jhonparra18/gpt2-med-sft-chat-spanish"
  
  model = AutoModelForCausalLM.from_pretrained(base_model,torch_dtype=torch.float16,device_map="auto")
  tokenizer = AutoTokenizer.from_pretrained(base_model)

  #define your generation args
  generation_kwargs={
      'num_beams':5,
      'no_repeat_ngram_size':2,
      'do_sample':True,
      'early_stopping':True,
      'top_p':0.95
  }

  tokenizer.padding_side="left"
  model_inputs=tokenizer.apply_chat_template(messages,tokenize=True,return_tensors="pt",return_dict=True,padding=True)
  outputs=model.generate(
      **model_inputs,
      **generation_kwargs
  )
  for i, output in enumerate(outputs):
    print("{}: {}".format(i, tokenizer.decode(output, skip_special_tokens=True)))

  from transformers import pipeline

  pipe=pipeline("text-generation",model=base_model,device_map="auto",**generation_kwargs)
  outputs=pipe(messages)
  print(outputs[0]['generated_text'][-1]['content'])