search

Jiayuan-Gu / hab-mobile-manipulation

Mobile manipulation in Habitat
https://sites.google.com/view/hab-m3
62 stars 9 forks source link

How to train Monolithic Baseline? #9

Closed albzni closed 4 months ago

albzni commented 6 months ago

Hi @Jiayuan-Gu. Thanks for sharing such a great project implementation! I notice that in your paper you utilize Monolithic RL as the worst baseline comparison, may I ask how I can reproduce this baseline performance using your code (or how I can modify the contents of the config file)?

Jiayuan-Gu commented 6 months ago

Thank you for the interest in our work. I will try to update them in the next few weeks.

albzni commented 5 months ago

Hello @Jiayuan-Gu.

Thank you indeed for your prompt response. Your willingness to update the monolithic baseline codes is sincerely appreciated!

I wonder if you might share some insight into your timeframe for implementing these updates, without any intention to rush you. Having a rough idea when they might be available would be greatly beneficial for planning purposes.

Thank you once again for your time and aid!

ChangFengPoLang commented 5 months ago

I also want to know the answer to this question.

Jiayuan-Gu commented 5 months ago

Hi @albzni and @ChangFengPoLang Sorry for the late reply. I have uploaded the configs relevant to monolithic policies to the branch mono: https://github.com/Jiayuan-Gu/hab-mobile-manipulation/tree/mono. For the SetTable task, the monolithic policy can not handle the first stage (opening the drawer, picking the object), and thus I do not include later stages. Please let me know if you have any question.

albzni commented 4 months ago

Oh, that's great! Thank you very much for your reply! I will try to train the mono policies as soon as possible. : )