demo.md

To make it easier to get started, we've prepared a basic pipeline that guides users from data caching to data generation and visualization.

Usage steps:

1. Prepare the environment
2. Download the nuplan dataset
3. Download the model checkpoint from Huggingface (or your specified location).
4. Edit the update_config function in scripts/pipeline/pipeline.py to specify your nuplan data root, map root, and checkpoint path, for example:

   cfg.scenario_builder.data_root = '/path/to/your/nuplan/data'
   cfg.scenario_builder.map_root = '/path/to/your/nuplan/maps'
   cfg.checkpoint = '/path/to/your/model.ckpt'

5. Run the pipeline:

   python scripts/pipeline/pipeline.py

After running pipeline.py, you will see three kinds of visualization GIFs in the scripts/pipeline/ directory:

1. task_condition.gif - Shows the scene conditions and context

2. task_gt.gif - Shows the ground truth trajectories of all agents

3. task_pred.gif - Shows the model's predicted trajectories

No additional parameters are needed if the above paths are set correctly in the config.