# modelo.py

# Instalar bibliotecas necessárias
!pip install transformers datasets huggingface_hub

from datasets import load_dataset
from transformers import DistilBERTForSequenceClassification, DistilBERTTokenizer, Trainer, TrainingArguments
from huggingface_hub import HfApi

# Carregar o dataset IMDb
dataset = load_dataset('imdb')

# Carregar o tokenizer e o modelo
tokenizer = DistilBERTTokenizer.from_pretrained('distilbert-base-uncased')
model = DistilBERTForSequenceClassification.from_pretrained('distilbert-base-uncased')

# Tokenizar o dataset
def tokenize_function(examples):
    return tokenizer(examples['text'], padding='max_length', truncation=True)

tokenized_datasets = dataset.map(tokenize_function, batched=True)
tokenized_datasets = tokenized_datasets.remove_columns(['text'])
tokenized_datasets.set_format('torch')

# Configurar os argumentos de treinamento
training_args = TrainingArguments(
    output_dir='./results',
    evaluation_strategy='epoch',
    learning_rate=2e-5,
    per_device_train_batch_size=16,
    per_device_eval_batch_size=16,
    num_train_epochs=3,
    weight_decay=0.01,
)

# Criar o trainer
trainer = Trainer(
    model=model,
    args=training_args,
    train_dataset=tokenized_datasets['train'],
    eval_dataset=tokenized_datasets['test'],
)

# Treinar o modelo
trainer.train()

# Salvar o modelo
model.save_pretrained("imdb-distilbert")
tokenizer.save_pretrained("imdb-distilbert")

# Fazer login no Hugging Face (substitua 'seu-token' pelo seu token de acesso)
!huggingface-cli login --token seu-token

# Enviar o modelo para o Hugging Face
api = HfApi()
api.upload_folder(
    folder_path="imdb-distilbert",
    path_in_repo="",
    repo_id="seu-username/imdb-distilbert",
    repo_type="model"
)

print("Deploy completo! Acesse seu modelo no Hugging Face para mais detalhes.")