ACT trained on PushT dataset for 80,000 training steps. Trained for 3.5 hrs on T4 GPU on Colab free.

The goal is to use this to compare ACTs performance on PushT against diffusion policy particularly the aspect of action multimodality.

Training logs: https://api.wandb.ai/links/ramachandranaadarsh-indian-institute-of-technology-madras/7m9dpcgw

Action Chunking with Transformers (ACT) is an imitation-learning method that predicts short action chunks instead of single steps. It learns from teleoperated data and often achieves high success rates.

This policy has been trained and pushed to the Hub using LeRobot. See the full documentation at LeRobot Docs.

For a complete walkthrough, see the training guide. Below is the short version on how to train and run inference/eval:

Prefix the dataset repo with eval_ and supply --policy.path pointing to a local or hub checkpoint.