Please help.....We definitely need a best practice for this scenario, as we need to load weight and bias from huge numpy files (500M, or 5G), and distribute them to each worker for the inference.FYI:1. load huge file in worker: each worker can load t...
