denisyarats / dmc2gym Goto Github PK

OpenAI Gym wrapper for the DeepMind Control Suite

License: MIT License

Python 100.00%

openai-gym dm-control gym reinforcement-learning openai deepmind

dmc2gym's Introduction

OpenAI Gym wrapper for the DeepMind Control Suite.

A lightweight wrapper around the DeepMind Control Suite that provides the standard OpenAI Gym interface. The wrapper allows to specify the following:

Reliable random seed initialization that will ensure deterministic behaviour.
Setting from_pixels=True converts proprioceptive observations into image-based. In additional, you can choose the image dimensions, by setting height and width.
Action space normalization bound each action's coordinate into the [-1, 1] range.
Setting frame_skip argument lets to perform action repeat.

Instalation

pip install git+git://github.com/denisyarats/dmc2gym.git

Usage

import dmc2gym

env = dmc2gym.make(domain_name='point_mass', task_name='easy', seed=1)

done = False
obs = env.reset()
while not done:
  action = env.action_space.sample()
  obs, reward, done, info = env.step(action)

dmc2gym's People

Stargazers

Watchers

Forkers

fagan2888 geyang yusukeurakami mishalaskin amrmkayid krishnachaitanya7 anthliu yangrui2015 aravindr93 quanzhou-li frt03 nii4u mwulmer melfm ssilwal liutong966 arnaudfickinger zzyunzhi dyth sahandrez zhihanyang2022 yudasong cvoelcker yawen-d nicoguertler mimeku humancompatibleai matzewolf langep7 jan1854 c-rizz xvsfekcn lee15253 onnoeberhard simipixel ycl010203 velythyl naumix gumussoysuat kevingmelin prakyathkantharaju zjowowen hymwgk tohsin mooricanna ltzheng minoring haooooooqi frankroeder bowenxxxx jimimvp 5l1v3r1 zibindong borchand c0sch0 vbarbaros

dmc2gym's Issues

python: /builds/florianrhiem/pyGLFW/glfw-3.3.7/src/posix_thread.c:64: _glfwPlatformGetTls: Assertion `tls->posix.allocated == 1' failed.

Getting this error

python: /builds/florianrhiem/pyGLFW/glfw-3.3.7/src/posix_thread.c:64: _glfwPlatformGetTls: Assertion `tls->posix.allocated == 1' failed.

when it gets to the env.render() function. Not sure what to do.

import dmc2gym

env = dmc2gym.make(domain_name='point_mass', task_name='easy', seed=1)

done = False
obs = env.reset()
while not done:
  action = env.action_space.sample()
  obs, reward, done, info = env.step(action)
  rgb_obs = env.render(mode='rgb_array')

Unable to install dmc2gym

Thank you for open sourcing dmc2gym. I have been using dmc2gym since last year, but after yesterday I am unable to install dmc2gym using the mentioned installation command-
pip install git+git://github.com/denisyarats/dmc2gym.git

It is showing the error-

I checked the mentioned link in the error message and found that Github has disabled support for unencrypted Git protocol (git://). What is the alternative way to install dmc2gym?

Request your help in this matter @denisyarats @bamos

dm-control OpenGL display error

Hey Denis, thanks for this wrapper and congrats on your SAC_AE paper!

people have been running into a dm-control error when calling from dm_control import mujoco from within dmc2gym.

You can see the error message here: google-deepmind/dm_control#123 (comment)

Will post here if I find a solution.

Plans to switch to gymnasium

Since the original OpenAI gym is no longer maintained, do you have plans to move or upgrade to gymnasium? Or would it be better for that to be a different project due to the breaking API changes in gymnasium?

The new API is nicely improved especially for the continuous control tasks of DMC with the addition of the termination and truncation.

how can I visualize the game?

Installation problem

Hi,
Provided installation command isn't working properly. It seems that it's unable to connect to github.com. I found an alternative solution from here which worked for me. Simply, I tried this: "pip install git+https://github.com/1nadequacy/dmc2gym.git" (https instead of git)

time_limit = None RunTime error

When running the example, we got the following bug:

import dmc2gym

env = dmc2gym.make(domain_name='point_mass', task_name='easy', seed=1)

done = False
obs = env.reset()
while not done:
    action = env.action_space.sample()
    obs, reward, done, info = env.step(action)

Error

/Users/ge/opt/anaconda3/envs/plan2vec/bin/python /Users/ge/fair/dm_control_suite_playground/dm_control_suite_playground/__init__.py
Traceback (most recent call last):
  File "/Users/ge/fair/dm_control_suite_playground/dm_control_suite_playground/__init__.py", line 3, in <module>
    env = dmc2gym.make(domain_name='point_mass', task_name='easy', seed=1)
  File "/Users/ge/fair/dmc2gym/dmc2gym/__init__.py", line 48, in make
    return gym.make(env_id)
  File "/Users/ge/opt/anaconda3/envs/plan2vec/lib/python3.6/site-packages/gym/envs/registration.py", line 156, in make
    return registry.make(id, **kwargs)
  File "/Users/ge/opt/anaconda3/envs/plan2vec/lib/python3.6/site-packages/gym/envs/registration.py", line 101, in make
    env = spec.make(**kwargs)
  File "/Users/ge/opt/anaconda3/envs/plan2vec/lib/python3.6/site-packages/gym/envs/registration.py", line 73, in make
    env = cls(**_kwargs)
  File "/Users/ge/fair/dmc2gym/dmc2gym/wrappers.py", line 64, in __init__
    environment_kwargs=environment_kwargs
  File "/Users/ge/opt/anaconda3/envs/plan2vec/lib/python3.6/site-packages/dm_control/suite/__init__.py", line 114, in load
    environment_kwargs, visualize_reward)
  File "/Users/ge/opt/anaconda3/envs/plan2vec/lib/python3.6/site-packages/dm_control/suite/__init__.py", line 148, in build_environment
    env = domain.SUITE[task_name](**task_kwargs)
  File "/Users/ge/opt/anaconda3/envs/plan2vec/lib/python3.6/site-packages/dm_control/suite/point_mass.py", line 49, in easy
    physics, task, time_limit=time_limit, **environment_kwargs)
  File "/Users/ge/opt/anaconda3/envs/plan2vec/lib/python3.6/site-packages/dm_control/rl/control.py", line 80, in __init__
    self._physics.timestep() * self._n_sub_steps)
TypeError: unsupported operand type(s) for /: 'NoneType' and 'float'

Process finished with exit code 1

dmc2gym for robosuite environments

Is it possible to use dmc2gym wrappers for the robosuite environments listed here? Or any other way to adapt dmc2gym for these environments? Thank you.

https://robosuite.ai/docs/source/robosuite.environments.manipulation.html

dmc2gym has no attribute 'env_specs'

Traceback (most recent call last):
File "/home/yang/dmcontrol-generalization-benchmark/src/train.py", line 150, in
main(args)
File "/home/yang/dmcontrol-generalization-benchmark/src/train.py", line 50, in main
mode='train'
File "/home/yang/dmcontrol-generalization-benchmark/src/env/wrappers.py", line 48, in make_env
background_dataset_paths=paths
File "/home/yang/dmcontrol-generalization-benchmark/src/env/dmc2gym/dmc2gym/init.py", line 36, in make
if not env_id in gym.envs.registry.env_specs:
AttributeError: 'dict' object has no attribute 'env_specs'

pip install does not work

Hi
The install command mentioned here does not work.

pip install git+git://github.com/denisyarats/dmc2gym.git
Collecting git+git://github.com/denisyarats/dmc2gym.git
  Cloning git://github.com/denisyarats/dmc2gym.git to /tmp/pip-req-build-ahuyx_q5
  Running command git clone -q git://github.com/denisyarats/dmc2gym.git /tmp/pip-req-build-ahuyx_q5
  fatal: unable to connect to github.com:
  github.com[0: 140.82.121.4]: errno=Connection timed out

WARNING: Discarding git+git://github.com/denisyarats/dmc2gym.git. Command errored out with exit status 128: git clone -q git://github.com/denisyarats/dmc2gym.git /tmp/pip-req-build-ahuyx_q5 Check the logs for full command output.
ERROR: Command errored out with exit status 128: git clone -q git://github.com/denisyarats/dmc2gym.git /tmp/pip-req-build-ahuyx_q5 Check the logs for full command output.

Is there any other way I build the package and install it on my system? Any help will be useful.

Thanks & Regards
Swagat

from_pixels doesn't work

import dmc2gym

env = dmc2gym.make(domain_name='point_mass', task_name='easy', from_pixels=True)

done = False
obs = env.reset()
while not done:
  action = env.action_space.sample()
  obs, reward, done, info = env.step(action)

~/anaconda3/envs/main/lib/python3.7/site-packages/dmc2gym/__init__.py in make(domain_name, task_name, seed, visualize_reward, from_pixels, height, width, camera_id, frame_skip, episode_length, environment_kwargs, time_limit, channels_first)
     21 
     22     if from_pixels:
---> 23         assert not visualize_reward, 'cannot use visualize reward when learning from pixels'
     24 
     25     # shorten episode length

AssertionError: cannot use visualize reward when learning from pixels

How can I fix it?

Any plans of publishing this to pypi?

Hi @denisyarats, thank you for this amazing project. Any plans of publishing this to pypi yet?