from datasets import load_dataset
from transformers import AutoTokenizer, AutoModelForSequenceClassification, Trainer, TrainingArguments
import numpy as np
from sklearn.metrics import accuracy_score, precision_recall_fscore_support

model_name = "indobenchmark/indobert-base-p1"

dataset = load_dataset("csv", data_files="data/eval_dataset.csv")

label_map = {"negative":0, "neutral":1, "positive":2}

def encode_label(example):
    example["label"] = label_map[example["label"]]
    return example

dataset = dataset.map(encode_label)

tokenizer = AutoTokenizer.from_pretrained(model_name)

def preprocess(example):
    return tokenizer(example["text"], truncation=True, padding=True)

dataset = dataset.map(preprocess)

model = AutoModelForSequenceClassification.from_pretrained(model_name, num_labels=3)

def compute_metrics(eval_pred):
    logits, labels = eval_pred
    preds = np.argmax(logits, axis=1)

    precision, recall, f1, _ = precision_recall_fscore_support(labels, preds, average='weighted')
    acc = accuracy_score(labels, preds)

    return {
        "accuracy": acc,
        "f1": f1,
        "precision": precision,
        "recall": recall
    }

training_args = TrainingArguments(
    output_dir="./model",
    learning_rate=2e-5,
    per_device_train_batch_size=8,
    num_train_epochs=3,
    evaluation_strategy="epoch"
)

trainer = Trainer(
    model=model,
    args=training_args,
    train_dataset=dataset["train"],
    eval_dataset=dataset["train"],
    compute_metrics=compute_metrics
)

trainer.train()

# 🔥 simpan model
trainer.save_model("model/final_model")
tokenizer.save_pretrained("model/final_model")