Running the agent in different sensor settings as in the ablation study

I would like to ask about running the agent in different sensor settings as shown in Table 2 of your paper.
Do we need to generate data for each sensor setup and retrain the agent?
Or do you provide your pre-trained models for the different settings?
