v0.1.0 Initial Release

• Initial release
• Implements federated averaging using an arbitrary number of clients
• Implements a non-federated learning baseline to which federated learning algorithms can be compared
• Supports client sub-sampling, i.e., only a subset of all clients participates in each communication round
• Extensively logs hyperparameters and training statistics
• Intelligently retains model checkpoint files
• Training statistics can be plotted
• Supports the following models:
  • LeNet-5
  • VGG11
• Supports the following datasets:
  • MNIST
  • CIFAR-10
• Supports Linux on AMD64 and MacOS on ARM64
• Performed extensive experiments on all supported models and datasets using various numbers of clients and recorded the results in the read me