Error Training MyoLeg Agent with depRL

[nikypopov]

Issue: Error Training MyoLeg Agent with depRL

Hello!

I'm running into an error when trying to train the MyoLeg agent with depRL. Specifically, following this example

Steps to Reproduce

1. Attempt to Train Agent:

   python -m deprl.main baselines_DEPRL/myoLegWalk.json
   

   Error:

   raise error.NameNotFound(
   gym.error.NameNotFound: Environment myoLegWalk doesn't exist.
   

I was able to reproduce the same error when just attempting to load a pre-trained depRL model with the following script:

2. Load depRL Baseline for Walking:

   import gym
   import myosuite
   import deprl
   
   # we can also change the reset_type of the environment here
   env = gym.make('myoLegWalk-v0', reset_type='random')
   policy = deprl.load_baseline(env)
   
   for ep in range(5):
       obs, _ = env.reset()
       for i in range(1000):
           action = policy(obs)
           next_obs, reward, done, truncated, info = env.step(action)
           env.sim.renderer.render_to_window()
           obs = next_obs
           if done:
               break
   

   Error:

   raise error.NameNotFound(
   gym.error.NameNotFound: Environment myoLegWalk doesn't exist.
   

And this error could be resolved changing the import statement for gym to:

from myosuite.utils import gym

This does fix the issue in the above example script but does not resolve the training error.

3. Attempted to Edit .json configuration file: I also tried changing the header tab in the configuration myoLegWalk.json file to:

 "tonic": {
     "header": "import deprl, myosuite\nfrom myosuite.utils import gym",   

however this didn't fix the issue and returned the same error.

I'm wondering if this has anything to do with a myosuite-gym (or gymnaisum) update? Any help would be greatly appreciated, thanks!

Configuration and Environment

MacOS Configuration:

• MacOS 14.4.1
• Python 3.8.19
• depRL 0.3.2
• MyoSuite 2.4.0

Linux Configuration:

• Python 3.9.19
• depRL 0.4.0
• MyoSuite 2.4.0

On Linux, attempting to train the MyoLeg agent yielded:

raise error.UnregisteredEnv('No registered env with id: {}'.format(id)) gym.error.UnregisteredEnv: No registered env with id:  myoLegWalk-v0