Optimizing the batch size of training ought to be a server because it has to optimize across multiple recurrences of the job. Then, BatchSizeOptimizer (under zeus.optimizer.batch_size) will act as the client for the server, and will be integrated into the user's training script.
Optimizing the batch size of training ought to be a server because it has to optimize across multiple recurrences of the job. Then,
BatchSizeOptimizer(underzeus.optimizer.batch_size) will act as the client for the server, and will be integrated into the user's training script.