Dataset

Please get dataset from this link and put in human-demo/can-pick: https://drive.google.com/drive/folders/1upTXzwI3AsTnKlqR_mz6qz_1J0boQ9I3?usp=sharing

Training

To run the training, run

experiments/experiment_training.py

add

--exp_name=your_name_here

to log experiment in different names

How to change reward shaping

add item to

shape_reward = []

line 74 in

experiments/experiment_training.py

currenly supports: "progress_sign_loss","delta_progress_scale_loss", "value_sign_loss", "advantage_sign_loss" , "progress_value_loss"