Language Processing (LP) component

[SeungyounShin]

We were reviewing the main.py script that loads traj_data from alfred_data_all, a subset of the data, and generates evaluation results for multiple modules, excluding the BERT module. However, upon submitting the JSON file to the leaderboard, we noticed that it does not include the low-level instruction data produced by the LP component. We would greatly appreciate it if you could provide us with some clarification on this matter.
Additionally, we were wondering if alfred_data_all is pre-computed with the BERT module

[soyeonm]

Hi,

I think you can use the "--appended" flag when running the command to control this.
"--appended": use high + low level
without this flag: use only high level

But please let me know if your concern is not resolved by this!