0.4.0 (2024-11-29)

• Better logging of meta information during training
• Use size of input data to construct model
• Do not encode image size into model parameters (for better flexibility)