注册OpenAI Gym Environment时环境不存在

Question

I am trying to register a custom gym environment on a remote server, but it is not working. I have been able to successfully register this environment on my personal computer using the Anaconda package manager framework, but have so far been unsuccesful without Anaconda (so I know the problem is not my environment). To try and work through this problem, I followed this tutorial .

Here is my file structure:

gym-knot
│   setup.py  
│
└───gym_knot
    │   __init__.py
    │   
    └───envs
        │   __init__.py
        │   sliceenv.py

Where gym-knot/setup.py is

from setup tools import setup

setup(name='gym_knot', version='0.0.1', install_requires=['gym'] 

gym-knot/gym_knot/__init__.py is

from gym.envs.registration import register

register(
   id='Slice-v0',
   entry_point='gym_knot.envs:SliceEnv',
   max_episode_steps=300,
   reward_threshold=0,)

gym-knot/gym_knot/envs/__init__.py is

from gym_knot.envs.sliceenv import SliceEnv

I then ran pip install --user -e gym-knot to install it.

I have import gym_knot in the python file with my algorithm.

When I use the command python3 knot_project.py (which is the python file with my algorithm and whatnot), I get the error:

gym.error.NameNotFound: Environment 'gym_knot:Slice' doesn't exist.

I am really confused why I am getting this error because I followed the tutorial exactly and have also look at other resources that told me the same thing. Am I missing something? Could it be because my knot_project.py file is not in the gym-knot folder so it does not know that my environment exists? Any help would be appreciated.

Answer 1

This started happening to me after I upgraded gym from version 0.21 to 0.23, no solution yet but a good clue

[Update] It looks like it was a broken update

Download the zip file for 0.24.0 (or higher) for OpenAI from https://github.com/openai/gym/releases and manually install this version with pip install path/to/dir

See Re-added gym.make("MODULE:ENV") import style that was accidentally removed in v0.22 @arjun-kg comment on patch notes