This list describes the planned features including breaking changes.
- Benchmark MuJoCo datasets
- Benchmark Atari 2600 datasets
- Change MDPDataset format to align with D4RL datasets
- Sophisticated config system using dataclasses
- Dump configuration and model parameters in a single file
- Support large dataset
- Support tuple observation
- Support large-scale data-parallel offline training
- Support Transformer architecture (e.g. Decision Transformer)
- Speed up training with CudaGraph and torch.compile
- Support training foundation models (e.g. Gato)
- Support large-scale distributed online training
- Change library name to represent unification of offline and online