import gradio as gr

# Load model directly
from transformers import AutoTokenizer, AutoModelForSeq2SeqLM

tokenizer = AutoTokenizer.from_pretrained("yeye776/t5-OndeviceAI-HomeIoT")
model = AutoModelForSeq2SeqLM.from_pretrained("yeye776/t5-OndeviceAI-HomeIoT")

# Gradio 인터페이스 구성
def generate_answer(input_text):
    # 입력 텍스트를 모델 토크나이저로 토큰화
    input_ids = tokenizer(input_text, max_length=700, return_tensors="pt").input_ids
    
    # 모델 추론
    output_ids = model.generate(input_ids, top_k=10, max_length=1024)
    # output_ids = model.generate(input_ids, num_beams=10, top_k=10, max_length=1024)
    
    # 모델 출력을 텍스트로 디코딩
    output_text = tokenizer.decode(output_ids[0], skip_special_tokens=True)

    return output_text

# Gradio 인터페이스 정의
iface = gr.Interface(
    fn=generate_answer,
    inputs="text", 
    outputs="text",
    title="OnDevice & AI Home IoT",
    description="Hugging Face Transformers + Gradio Demo"
)

# Gradio 앱 시작
iface.launch()

# gr.load("models/yeye776/t5-OndeviceAI-HomeIoT").launch()
# iface = gr.Interface(fn=pipe, inputs="text", outputs="text")
# iface.launch()