how to train vision-based policy

[qiaojunyu]

Hi, Thanks for the amazing work!
We have successfully trained a state-based RL model on drawer and would like guidance on generating operation data based on point clouds and further training the vision-based policy. We would appreciate your expert advice on the recommended approach for achieving this.

Thank you for your assistance.

[geng-haoran]

Will release as soon as possible after testing!

[jaredmejia]

@geng-haoran Any update on this?

[geng-haoran]

We put the original code in the branch "not_polish". We will refine them as soon as possible to fit into our existing released framework. check it at PartManip/not-polished_code in branch "not_polish".