search

pat-coady / trpo

Trust Region Policy Optimization with TensorFlow and OpenAI Gym
https://learningai.io/projects/2017/07/28/ai-gym-workout.html
MIT License
360 stars 106 forks source link

able to run FetchPickAndPlace-v1 ? #28

Closed MrDadaGuy closed 4 years ago

MrDadaGuy commented 5 years ago

Trying to get some of the Openai gym Robotics environments working. I've added the appropriate imports so it pulls in FetchPickAndPlace-v1 but the script fails on this line

obs_dim = env.observation_space.shape[0]

I've had a look and shape is None -- although observation_space looks like this:

Dict(achieved_goal:Box(3,), desired_goal:Box(3,), observation:Box(25,))

any ideas how I could modify the code to get past this?

Thanks!

MrDadaGuy commented 5 years ago

I've hacked basic support for OpenAI Gym Robotics in my fork...

https://github.com/MrDadaGuy/trpo

I'm running FetchPickAndPlace-v1 now

pat-coady commented 4 years ago

I've migrated the repo over to PyBullet (and TensorFlow2). I don't have a MuJoCo license any more, so am unable to test test FetchPickAndPlace-v1. I fixed some things related to the way action and observation shapes were handled. Hopefully this resolves the problem with that env.