QuickStart | Development | Results
RayCarla is a large scale autonomous driving research platform.
- Simulation: Carla simulator.
- Imitative Learning on Ray to support distributed deep learning.
- Reinforcement Learning on Ray (WIP).
- Carla on Ray to support distributed simulation (WIP).
- Ray Serve to support distributed inference (WIP).
- Leverage Ray Dataset for distributed data ingestion (WIP).
Follow raycarla.ipynb notebook to enjoy!
Please follow DEVELOPMENT.md to setup environment.
The training time has been reduced from 20 hours to 3 hours after Ray enale the training on 8 GPUs in parallel.