1️ Configurar hiperparâmetros e treinar o modelo
Usamos a biblioteca Transformers da Hugging Face .
hiperparâmetros
from transformers import TrainingArguments
training_args = TrainingArguments(
output_dir="./modelo-redacao",
learning_rate=2e-5,
per_device_train_batch_size=16,
per_device_eval_batch_size=16,
num_train_epochs=3,
evaluation_strategy="epoch",
save_strategy="epoch"
)
otimizador
from transformers import Trainer
trainer = Trainer(
model=model,
args=training_args,
train_dataset=train_dataset,
eval_dataset=test_dataset,
tokenizer=tokenizer
)
Treinar o modelo
trainer.train()
