app.py

import gradio as gr
import pandas as pd
from datasets import Dataset
from transformers import AutoTokenizer, AutoModelForSequenceClassification, TrainingArguments, Trainer
from sklearn.metrics import accuracy_score, f1_score
import torch
import os

model_name = "distilbert-base-multilingual-cased"
tokenizer = AutoTokenizer.from_pretrained(model_name)

def preprocess_function(examples):
    return tokenizer(examples["comment"], truncation=True, padding=True)



def compute_metrics(eval_pred):
    logits, labels = eval_pred
    predictions = torch.argmax(torch.tensor(logits), dim=-1)
    acc = accuracy_score(labels, predictions)
    f1 = f1_score(labels, predictions, average="weighted")
    return {"accuracy": acc, "f1": f1}

def train_model(file):
    try:
        df = pd.read_csv(file.name)
        if "comment" not in df.columns or "label" not in df.columns:
            return "File CSV phải có cột 'comment' và 'label'"
    
        dataset = Dataset.from_pandas(df)
        tokenized_dataset = dataset.map(preprocess_function, batched=True)
    
        tokenized_dataset = tokenized_dataset.train_test_split(test_size=0.2)
        train_dataset = tokenized_dataset["train"]
        eval_dataset = tokenized_dataset["test"]
    
        model = AutoModelForSequenceClassification.from_pretrained(model_name, num_labels=2)
    
        args = TrainingArguments(
            output_dir="results",
            evaluation_strategy="epoch",
            save_strategy="epoch",
            learning_rate=2e-5,
            per_device_train_batch_size=8,
            per_device_eval_batch_size=8,
            num_train_epochs=3,
            weight_decay=0.01,
            logging_dir="logs",
            logging_steps=10,
            push_to_hub=True,
            hub_model_id="vnanhtuan/fine-tune-danh-gia-cam-xuc"
        )
    
        trainer = Trainer(
            model=model,
            args=args,
            train_dataset=train_dataset,
            eval_dataset=eval_dataset,
            tokenizer=tokenizer,
            compute_metrics=compute_metrics,
        )
    
        trainer.train()
        trainer.push_to_hub()
        return "Huấn luyện hoàn tất và model đã được đẩy lên Hugging Face."
    except Exception as e:
        import traceback
        return f"❌ Error: {str(e)}\n\n{traceback.format_exc()}"

def predict_sentiment(text):
    model = AutoModelForSequenceClassification.from_pretrained("vnanhtuan/fine-tune-danh-gia-cam-xuc")
    inputs = tokenizer(text, return_tensors="pt", truncation=True, padding=True)
    with torch.no_grad():
        outputs = model(**inputs)
    prediction = torch.argmax(outputs.logits, dim=1).item()
    label_text = "Tích cực" if prediction == 1 else "Tiêu cực"
    return label_text

with gr.Blocks() as demo:
    gr.Markdown("# Fine-tune mô hình phân loại cảm xúc tiếng Việt")

    with gr.Tab("Huấn luyện mô hình"):
        csv_file = gr.File(label="Tải file CSV gồm 2 cột: comment, label")
        train_button = gr.Button("Bắt đầu huấn luyện")
        train_output = gr.Textbox(label="Kết quả")

        train_button.click(fn=train_model, inputs=csv_file, outputs=train_output)

    with gr.Tab("Dự đoán cảm xúc"):
        input_text = gr.Textbox(label="Nhập câu cần phân tích")
        output_label = gr.Textbox(label="Kết quả dự đoán")
        predict_button = gr.Button("Dự đoán")

        predict_button.click(fn=predict_sentiment, inputs=input_text, outputs=output_label)

demo.launch()