src/warmup.py

import os

# Disable GPU and CPU-based MKL-DNN optimization to prevent container segmentation faults
os.environ["CUDA_VISIBLE_DEVICES"] = "-1"
os.environ["FLAGS_use_mkldnn"] = "0"
os.environ["FLAGS_use_xdnn"] = "0"
os.environ["OMP_NUM_THREADS"] = "1"

from paddleocr import PaddleOCR
from src.config import settings

def warmup_models():
    print("Starting warmup: pre-downloading PaddleOCR models...")
    for lang in settings.SUPPORTED_LANGUAGES:
        print(f"Downloading/Caching core model: '{lang}'...")
        # Instantiating the object triggers the download
        PaddleOCR(use_angle_cls=True, lang=lang, show_log=False, enable_mkldnn=False)
    print("Warmup complete! All core models are successfully cached.")

if __name__ == "__main__":
    warmup_models()