I am trying to generate a multi-task dataset using the provided dataset_generator script. However, when I load the low_dim_obs.pkl, some timesteps seem to miss the _observations.misc["executed_demo_joint_position_action"] property, which makes the trajectories discontinuous. At first I thought it might be just the first and the last timesteps, however, the missing values are often in the middle of the trajectory. Is this a known bug or am I missing some rationale behind this engineering decision.
Hi,
I am trying to generate a multi-task dataset using the provided dataset_generator script. However, when I load the low_dim_obs.pkl, some timesteps seem to miss the _observations.misc["executed_demo_joint_position_action"] property, which makes the trajectories discontinuous. At first I thought it might be just the first and the last timesteps, however, the missing values are often in the middle of the trajectory. Is this a known bug or am I missing some rationale behind this engineering decision.
Thanks