Skip to content

Latest commit

History

History
33 lines (24 loc) 路 1.14 KB

File metadata and controls

33 lines (24 loc) 路 1.14 KB

Paper

https://diffusion-policy.cs.columbia.edu

Training

The reference implementation maintains an exponential moving average (EMA) of the policy weights during training and evaluates the EMA weights. To reproduce this behavior, enable the trainer's EMA shadow:

lerobot-train \
  --policy.type=diffusion \
  --ema.enable=true \
  ...

Checkpoints then contain a directly loadable copy of the EMA weights next to the live ones, e.g. for evaluation:

lerobot-eval --policy.path=outputs/train/.../checkpoints/last/pretrained_model_ema ...

The EMA decay schedule (--ema.inv_gamma, --ema.power, ...) defaults to the reference implementation's values. For a constant decay instead of the warmup schedule (e.g. to match openpi's pi0/pi05 training), set --ema.decay=0.99.

Citation

@article{chi2024diffusionpolicy,
	author = {Cheng Chi and Zhenjia Xu and Siyuan Feng and Eric Cousineau and Yilun Du and Benjamin Burchfiel and Russ Tedrake and Shuran Song},
	title ={Diffusion Policy: Visuomotor Policy Learning via Action Diffusion},
	journal = {The International Journal of Robotics Research},
	year = {2024},
}