I think this is super-cool, especially when juxtaposed with the popularity of super gigantic models that easily take thousands of dollars of compute to train.
The homepage was helpful in delineating the specific use case (when compared to `torch.rpc`, `torch.elastic`, ...etc).
The homepage was helpful in delineating the specific use case (when compared to `torch.rpc`, `torch.elastic`, ...etc).
Any words on how this compares to TensorFlow's Federating Learning offerring? https://www.tensorflow.org/federated/federated_learning