app.py

import gradio as gr
from fastapi import FastAPI, HTTPException, Request
from pydantic import BaseModel
from transformers import AutoModelForCausalLM, AutoTokenizer
import torch
import uvicorn
from fastapi.responses import HTMLResponse

# Khởi tạo FastAPI
app = FastAPI()

# Tải model và tokenizer khi ứng dụng khởi động
model_name = "Qwen/Qwen2.5-0.5B"
try:
    tokenizer = AutoTokenizer.from_pretrained(model_name)
    model = AutoModelForCausalLM.from_pretrained(
        model_name,
        torch_dtype="auto",
        device_map="auto",
        attn_implementation="eager"  # Tránh cảnh báo sdpa
    )
    print("Model and tokenizer loaded successfully!")
except Exception as e:
    print(f"Error loading model: {e}")
    raise

# Định nghĩa request body cho API
class TextInput(BaseModel):
    prompt: str
    max_length: int = 100

# Hàm sinh văn bản (dùng chung cho API và Gradio)
def generate_text(prompt, max_length=100):
    try:
        inputs = tokenizer(prompt, return_tensors="pt").to(model.device)
        outputs = model.generate(
            inputs["input_ids"],
            max_length=max_length,
            num_return_sequences=1,
            no_repeat_ngram_size=2,
            do_sample=True,
            top_k=50,
            top_p=0.95
        )
        return tokenizer.decode(outputs[0], skip_special_tokens=True)
    except Exception as e:
        raise Exception(f"Error: {str(e)}")

# API endpoint để sinh văn bản
@app.post("/generate")
async def generate_text_api(input: TextInput):
    try:
        result = generate_text(input.prompt, input.max_length)
        return {"generated_text": result}
    except Exception as e:
        raise HTTPException(status_code=500, detail=str(e))

# API endpoint kiểm tra sức khỏe
@app.get("/")
async def root():
    return {"message": "Qwen2.5-0.5B API is running!"}

# API endpoint hiển thị URL
@app.get("/api_link")
async def get_api_link(request: Request):
    scheme = request.url.scheme
    host = request.url.hostname
    if request.url.port:
        base_url = f"{scheme}://{host}:{request.url.port}"
    else:
        base_url = f"{scheme}://{host}"
    return {
        "api_url": base_url,
        "endpoints": {
            "health_check": f"{base_url}/",
            "generate_text": f"{base_url}/generate",
            "api_link": f"{base_url}/api_link",
            "interface": f"{base_url}/interface"
        }
    }

# Tạo giao diện Gradio
def create_gradio_interface():
    with gr.Blocks(title="Qwen2.5-0.5B Text Generator") as demo:
        gr.Markdown("# Qwen2.5-0.5B Text Generator")
        gr.Markdown("Enter a prompt and get generated text!")
        
        with gr.Row():
            prompt_input = gr.Textbox(label="Prompt", placeholder="Type something...")
            max_length_input = gr.Slider(50, 500, value=100, step=10, label="Max Length")
        
        generate_button = gr.Button("Generate")
        output_text = gr.Textbox(label="Generated Text", interactive=False)
        
        generate_button.click(
            fn=generate_text,
            inputs=[prompt_input, max_length_input],
            outputs=output_text
        )
    return demo

# Thêm endpoint để hiển thị giao diện Gradio
@app.get("/interface", response_class=HTMLResponse)
async def gradio_interface(request: Request):
    gradio_app = create_gradio_interface()
    return HTMLResponse(content=gradio_app.render())

# Chạy ứng dụng nếu không trên Hugging Face Spaces
if __name__ == "__main__":
    uvicorn.run(app, host="0.0.0.0", port=7860)