Closed ssvision closed 2 months ago
Hi, we are working on adding several models to the library, namely:
We'll also add more documentation & tutorials on how to switch tasks when training a given policy ;)
@ssvision if you train a model for a policy/dataset/env combination we haven't already, and it matches or exceeds a SOTA benchmark, we can add it on the hub! What are you mos interested in? I heard aloha insertion. The reason we haven't added that is that (a) the results are pretty poor on the ACT paper with only 20% success rate, (b) we actually haven't been able to reproduce, falling about 5% short either because of not trying enough seeds or maybe some of these tweaks that still need to go into ACT: https://github.com/huggingface/lerobot/pull/265
Closing due to inactivity. @ssvision if you have further questions, please don't hesitate to re-open.
how do i train and visualise my policy in the environment for other aloha and diffusion tasks? the hugging face repo has only 2 models namely diffusion_pusht & act_aloha_sim_transfer_cube_human?
how do i do the same for other tasks such as aloha insertion task ? and other range of tasks