If you want to run the Docker image in a complete offline environment, you need to add the `--build-arg with_models=true` parameter. Then the language models are downloaded during the build process of the image. Otherwise these models get downloaded on the first run of the image/container.