TRL documentation
Usar el modelo después del entrenamiento
Usar el modelo después del entrenamiento
Once you have trained a model using either the SFTTrainer or DPOTrainer, you will have a fine-tuned model that can be used for text generation. In this section, we’ll walk through the process of loading the fine-tuned model and generating text. If you need to run an inference server with the trained model, you can explore libraries such as text-generation-inference.
Cargar y generar
Si has realizado un ajuste fino completo de un modelo, es decir, sin usar PEFT, puedes simplemente cargarlo como cualquier otro modelo de lenguaje en transformers:
import torch
from transformers import AutoTokenizer, AutoModelForCausalLM
model_name_or_path = "Qwen/Qwen3-0.6B" #path/to/your/model/or/name/on/hub
device = torch.accelerator.current_accelerator() if torch.accelerator.is_available() else "cpu"
model = AutoModelForCausalLM.from_pretrained(model_name_or_path).to(device)
tokenizer = AutoTokenizer.from_pretrained(model_name_or_path)
inputs = tokenizer.encode("This movie was really", return_tensors="pt").to(device)
outputs = model.generate(inputs)
print(tokenizer.decode(outputs[0]))Alternativamente, también puedes usar el pipeline:
from transformers import pipeline
model_name_or_path = "Qwen/Qwen3-0.6B" #path/to/your/model/or/name/on/hub
pipe = pipeline("text-generation", model=model_name_or_path)
print(pipe("This movie was really")[0]["generated_text"])Usar adaptadores PEFT
from peft import PeftConfig, PeftModel
from transformers import AutoModelForCausalLM, AutoTokenizer
base_model_name = "Qwen/Qwen3-0.6B" #path/to/your/model/or/name/on/hub
adapter_model_name = "path/to/my/adapter"
model = AutoModelForCausalLM.from_pretrained(base_model_name)
model = PeftModel.from_pretrained(model, adapter_model_name)
tokenizer = AutoTokenizer.from_pretrained(base_model_name)También puedes fusionar los adaptadores en el modelo base para poder usar el modelo como un modelo de transformers normal, aunque el checkpoint será significativamente más grande:
model = AutoModelForCausalLM.from_pretrained(base_model_name)
model = PeftModel.from_pretrained(model, adapter_model_name)
model = model.merge_and_unload()
model.save_pretrained("merged_adapters")Una vez que el modelo esté cargado y los adaptadores fusionados o mantenidos por separado encima, puede ejecutar la generación como con un modelo normal, tal como se describe arriba.
Update on GitHub