🧩 Extensibility

Adapt your model/dataset once, train almost any combination of "model" × "dataset" × "loss function".

🔧 Maintainability

No need to worry about breaking the training codes of existing models/datasets/loss functions when adding new ones.

🚀 Efficiency

Multi-GPU parallel training; Auto batch size reduction to prevent CUDA OOM, Auto hyperparameter searching, Python built-in logger; structured experimental result saving (json)...