search

danijar / daydreamer

DayDreamer: World Models for Physical Robot Learning
https://danijar.com/daydreamer
261 stars 27 forks source link

How to run it on my robot? #6

Open defrag-bambino opened 1 year ago

defrag-bambino commented 1 year ago

Hi, awesome project - thanks for sharing your code.

I would love to try the algorithm on the hexapod I recently build. What are the necessary steps to do that? Create a custom Gym wrapper? Is there a guideline on how to find the correct config for such a custom setup/robot?

Thanks

danijar commented 1 year ago

Hi @defrag-bambino, I would recommend to use DreamerV3 directly, using the small model config and with --run.script parallel --run.train_ratio -1. You don't need to specify anything else, but you can explicitly specify the CNN/MLP inputs in the config if you don't want the agent to look at all observation keys.

You can return your env from the make_env() function in train.py. The env should either directly implement the Embodied API or Gym/DMEnv and then use the FromGym or FromDM wrappers to turn them into Embodied envs.

rex5679 commented 1 year ago

Hi. I'm trying to implement the algorithm in my UR5 arm.Can you please let me know the steps for the same? Thanks