How to get a video that contains IMAGINING with evaluate.py?

wayveai / mile

PyTorch code for the paper "Model-Based Imitation Learning for Urban Driving".

MIT License

365 stars 35 forks source link

How to get a video that contains IMAGINING with evaluate.py? #38

Closed airobotis closed 10 months ago

airobotis commented 10 months ago

How to get a video that contains IMAGINING with evaluate.py ? like Predicted scenario 1 (green light) video on https://wayve.ai/thinking/learning-a-world-model-and-a-driving-policy/

Do you offer any parameters to record it ?

Thanks a lot.

anthonyhu commented 10 months ago

Hello, this is achieved by setting the parameter is_dreaming=True in here https://github.com/wayveai/mile/blob/main/mile/models/mile.py#L397. This will ensure the model gets no additional image observation and will need to update its future state from its internal world model.

airobotis commented 10 months ago

I appreciate your quick and kind response : )