This model, from the paper Building reliable sim driving agents by scaling self-play, is a pre-trained policy for a reliable simulated driving agent. It was trained using self-play on the Waymo Open Motion Dataset and achieves a 99.8% goal completion rate with minimal undesired outcomes.

Project page: https://sites.google.com/view/reliable-sim-agents

Library: https://github.com/Emerge-Lab/gpudrive

To load a pre-trained policy, use the following:

from gpudrive.networks.late_fusion import NeuralNet

# Load pre-trained model via huggingface_hub
agent = NeuralNet.from_pretrained("daphne-cornelisse/policy_S10_000_02_27")

See tutorial 04 for all the details.

Downloads last month
314
Safetensors
Model size
51.2k params
Tensor type
F32
·
Video Preview
loading