Rework run_federated.py to allow running in a multimachine setting.
- Migrate the train dataset preprocessing function into the iterative process, rather than part of dataset building the client data function. This is necessary to enable serialization of datasets for sending to remote workers. PiperOrigin-RevId: 322691112
Showing
- tensorflow_federated/python/research/optimization/stackoverflow/BUILD 1 addition, 0 deletions...ederated/python/research/optimization/stackoverflow/BUILD
- tensorflow_federated/python/research/optimization/stackoverflow/run_federated.py 29 additions, 9 deletions...thon/research/optimization/stackoverflow/run_federated.py
Please register or sign in to comment