api.py

import os
import zipfile
import gdown
import torch
from fastapi import FastAPI, HTTPException
from pydantic import BaseModel
from transformers import AutoTokenizer, MT5ForConditionalGeneration
import uvicorn
from fastapi.middleware.cors import CORSMiddleware


ZIP_PATH = "./model.zip"
MODEL_PATH = "./saved_openie_model"


def download_and_extract_model():
    if not os.path.exists(MODEL_PATH) or not os.listdir(MODEL_PATH):
        print("جاري تحميل النموذج من Google Drive...")
        
        url = f"https://drive.google.com/uc?id=1MZHAeCaQAyyfi6b2Dh3V_JCdTe8fTVmP"
        gdown.download(url, ZIP_PATH, quiet=False)
        
        print("جاري فك ضغط النموذج...")
        with zipfile.ZipFile(ZIP_PATH, 'r') as zip_ref:
    
            zip_ref.extractall(".")
        if os.path.exists(ZIP_PATH):
            os.remove(ZIP_PATH)
            
        print(" تم تجهيز النموذج محلياً بنجاح!")
    else:
        print(" النموذج موجود مسبقاً، تخطي مرحلة التحميل.")


download_and_extract_model()


app = FastAPI(title="mT5 Relation Extraction API", version="1.0")
app.add_middleware(
    CORSMiddleware,
    allow_origins=["*"],
    allow_credentials=True,
    allow_methods=["*"],
    allow_headers=["*"],
)
device = torch.device("cpu")


print(" جاري تحميل النموذج والمُرمّز في الذاكرة...")
try:
    tokenizer = AutoTokenizer.from_pretrained(MODEL_PATH)
    model = MT5ForConditionalGeneration.from_pretrained(MODEL_PATH)
    model.to(device)
    model.eval() # وضع الاستنتاج
    print("تم تحميل النموذج بنجاح! الخادم جاهز لاستقبال الطلبات.")
except Exception as e:
    print(f" حدث خطأ أثناء تحميل النموذج: {e}")


class RelationRequest(BaseModel):
    text: str


@app.post("/extract_relation")
async def extract_relation_api(request: RelationRequest):
    try:
        prompt = "استخراج العلاقة: " + request.text 
        
        inputs = tokenizer(
            prompt, 
            return_tensors="pt", 
            max_length=64, 
            padding="max_length", 
            truncation=True
        )
        
        with torch.no_grad():
            outputs = model.generate(
                input_ids=inputs["input_ids"].to(device),
                attention_mask=inputs["attention_mask"].to(device),
                max_new_tokens=15,    
                num_beams=3,          
                early_stopping=True
            )
        
        predicted_relation = tokenizer.decode(outputs[0], skip_special_tokens=True).strip()
        
        return {"relation": predicted_relation}
    
    except Exception as e:
        raise HTTPException(status_code=500, detail=str(e))

if __name__ == "__main__":

    uvicorn.run(app, host="0.0.0.0", port=7860)