Is it good practice to save NLP Transformer based pre-trained models into file system in production environment

I have developed a multi label classifier using BERT. I’m leveraging Hugging Face Pytorch implementation for transformers.

I have saved the pretrained model into the file directory in dev environment. Now, the application is ready to be moved the production environment.

Is it a good practice to save the models into file system in prod ?
Can I serialize the model files and word embeddings into any DB and read again ?