You signed in with another tab or window. Reload to refresh your session.You signed out in another tab or window. Reload to refresh your session.You switched accounts on another tab or window. Reload to refresh your session.Dismiss alert
{{ message }}
This repository has been archived by the owner on Oct 31, 2023. It is now read-only.
This allows the code to first perform a center_cropper, then resize_shortest_edge, which means resize_shortest_edge has not played any role. This is completely opposite to the order in the 2022 settings file. If I train the agent based on the 2022 code, it will result in inconsistent performance in the 2023 code. The question I asked earlier also caused this #165. Can you please modify the provided settings file again?
The text was updated successfully, but these errors were encountered:
Sign up for freeto subscribe to this conversation on GitHub.
Already have an account?
Sign in.
Hello!
Have you paid attention to the execution order of image post-processing in the 2023 settings? The settings are as follows:
defaults:
- /habitat_baselines: habitat_baselines_rl_config_base
- /benchmark/nav/objectnav: objectnav_v2_hm3d_stretch_challenge
- /habitat_baselines/rl/policy/obs_transforms:
- center_cropper_base
- resize_shortest_edge_base
- _self_
This allows the code to first perform a center_cropper, then resize_shortest_edge, which means resize_shortest_edge has not played any role. This is completely opposite to the order in the 2022 settings file. If I train the agent based on the 2022 code, it will result in inconsistent performance in the 2023 code. The question I asked earlier also caused this #165. Can you please modify the provided settings file again?
The text was updated successfully, but these errors were encountered: