How to replicate the results of PPO trained by nuPlan dataset?
Closed this issue · 1 comments
Liuxueyi commented
Could you give the training command and requirements.txt? It is hard to replicate the training environment for me. And if you train the PPO agent using the Waymo dataset?
Looking forward to your reply, thanks.
QuanyiLi commented
Hi, sorry for the late reply. I didn't watch this repo until now...
To run the training scripts, follow these steps:
- Create a new conda env and make sure the python==3.8
- Clone ScenarioNet and MetaDrive, install them via
pip install -e .[train]
andpip install -e .
respectively - Run any of the training scripts under
scenarionet/scenarionet_training/scripts
Before this, you should setup the dataset correctly.