I do not think you can (and should not). Workers are separate processes that fork before they run any of your codes.
You can put the βmodelβ (what makes it big?) In the Redis database and access it for every employee. The best option is probably to create a separate service in which you run one instance and exchange data via HTTP or RPC from your working one (see nameko for a simple (micro) service.
Another option would be to use a single celery worker, as well as statistical calculations in the task.
source share