Contents of datasets.
'param': All the parameters in the policy network as a flattened vector.
'traj': prior trajectories in first 60 steps, as 's_0, a_0, a_1, a_2, s_3, a_3, a_4, a_5'.
'task': the success three states 's_m, s_{m+1}, s_{m+2}'
If you want to train with your dataset or task, you can privately design the trajectory dimensions and encode them to the same dimension (for example we used 128). You can use our pretrained model with the same behavior dimensions to finetune on your dataset.
Cite arxiv.org/abs/2407.10973
📝 Citation
If you find our model or dataset useful, please consider citing as follows:
@article{liang2024make,
title={Make-An-Agent: A Generalizable Policy Network Generator with Behavior-Prompted Diffusion},
author={Liang, Yongyuan and Xu, Tingqiang and Hu, Kaizhe and Jiang, Guangqi and Huang, Furong and Xu, Huazhe},
journal={arXiv preprint arXiv:2407.10973},
year={2024}
}