agrimgupta92 / metamorph Goto Github PK

View Code? Open in Web Editor NEW

104.0 104.0 15.0 40.6 MB

Code for "MetaMorph: Learning Universal Controllers with Transformers", Gupta et al, ICLR 2022

Dockerfile 1.07% Shell 17.58% Python 81.34%

metamorph's People

Contributors

Stargazers

Watchers

Forkers

zsoftwarerepository masterxiong riceroll adwaver4157 tmats ericonaldo madhu-korada linshuang22 georgegu lulu-yunze vineethkumar7 gamemachina nopenai mi-przystupa hyf1530588058

metamorph's Issues

Bug or bad usage?

When I use:
gym.make("Unimal-v0", agent_name="floor-1409-1-13-01-12-18-30"),
it says:

Traceback (most recent call last):

  File "<stdin>", line 1, in <module>

  File "/home/miniconda3/lib/python3.9/site-packages/gym/envs/registration.py", line 676, in make

    return registry.make(id, **kwargs)

  File "/home/miniconda3/lib/python3.9/site-packages/gym/envs/registration.py", line 520, in make

    return spec.make(**kwargs)

  File "/home/miniconda3/lib/python3.9/site-packages/gym/envs/registration.py", line 140, in make

    env = cls(**_kwargs)

  File "/home/app/metamorph/metamorph/envs/tasks/task.py", line 17, in make_env

    env = globals()[env_func](xml, agent_name, **kwargs)

  File "/home/app/metamorph/metamorph/envs/tasks/locomotion.py", line 89, in make_env_locomotion

    env.reset()

  File "/home/app/metamorph/metamorph/envs/tasks/unimal.py", line 158, in reset

    obs = self.reset_model()

  File "/home/app/metamorph/metamorph/envs/tasks/unimal.py", line 185, in reset_model

    observation = self._get_obs()

  File "/home/app/metamorph/metamorph/envs/tasks/unimal.py", line 81, in _get_obs

    obs.update(module.observation_step(self, self.sim))

  File "/home/app/metamorph/metamorph/envs/modules/agent.py", line 231, in observation_step

    limb_obs = self.get_limb_obs(sim)

  File "/home/app/metamorph/metamorph/envs/modules/agent.py", line 149, in get_limb_obs

    obs["body_idx"] = self._get_one_hot_body_idx()

  File "/home/app/metamorph/metamorph/envs/modules/agent.py", line 195, in _get_one_hot_body_idx

    one_hot_encoding[rows, body_idxs] = 1

IndexError: index 8 is out of bounds for axis 1 with size 8

Have tried many agents in the test dir, no one works.

Trained Model

Great work!
Could you provide a trained model shown in the demo?

Thanks.

How to use the code on windows system.

Thanks for the great work, however I am not able to run this code on windows system, can you provide some tutorial, including training and testing.

`manifest for nvidia` error

When I try setup by docker, the error occurred.

(simul) bak@bak:~/Projects/metamorph$ ./scripts/build_docker.sh
+ TAG=metamorph
+ PARENT=nvidia/cuda:11.1-cudnn8-devel-ubuntu20.04
+ INSTALL_SCRIPT=install_gpu_deps
++ id -u
+ USER_ID=1000
+ docker build -f docker/Dockerfile --build-arg PARENT_IMAGE=nvidia/cuda:11.1-cudnn8-devel-ubuntu20.04 --build-arg INSTALL_SCRIPT=install_gpu_deps --build-arg USER_ID=1000 -t metamorph .
Sending build context to Docker daemon  122.9MB
Step 1/35 : ARG PARENT_IMAGE
Step 2/35 : FROM $PARENT_IMAGE
manifest for nvidia/cuda:11.1-cudnn8-devel-ubuntu20.04 not found: manifest unknown: manifest unknown

Should I have to install cuda 11.1, cudnn8-devel version?

I created conda environment named simul. After the above error occurred, I installed cuda11.7 but it doesn't work.

`obs` is not a dictionary but only has str `'proprioceptive'`.

I got an error at class UnimalHeightObs(gym.ObservationWrapper), def observation(self, obs): from hfield.py

  File "/home/bak/Projects/metamorph/metamorph/envs/wrappers/hfield.py", line 211, in observation
    obs["torso_height"] = round(z_pos - 1, 3)
TypeError: 'str' object does not support item assignment

The def observation(self, obs): is called from class UnimalEnv(gym.Env): in unimal.py. Though the obs is a dictionary and has 'proprioceptive', 'edges' keys at the end of def reset(self):, obs is just a str 'proprioceptive' at def observation(self, obs): class UnimalHeightObs(gym.ObservationWrapper).

From the viewpoint of the train loop, the error occurs when the code creates (or initializes) the world environment. At ppo.py, obs cannot create at def train(self):.

To solve above issue, I added self.obs_temp = obs at def reset(self): at def reset(self): class UnimalEnv(gym.Env): and obs = self.obs_temp at class UnimalHeightObs(gym.ObservationWrapper), def observation(self, obs):to preserve obs information.

unimal.py

class UnimalEnv(gym.Env):
    """Superclass for all Unimal tasks."""

...........
...........
    ###########################################################################
    # Reset sim
    ###########################################################################
    def reset(self):
        if self.sim is None or cfg.ENV.NEW_SIM_ON_RESET:
            self.sim = self._get_sim()
        else:
            self.sim.reset()

        self.step_count = 0

        if self.viewer is not None:
            self.viewer.update_sim(self.sim)
        obs = self.reset_model()
        self._set_action_space()
        # This is just a temp initialization, the final obs space will depend
        # on SelectKeysWrapper
        self.observation_space = spu.convert_obs_to_space(obs)
        self.metadata["video.frames_per_second"] = int(np.round(1.0 / self.dt))
        self.obs_temp = obs  ### what I added

        return obs

hfield.py

class UnimalHeightObs(gym.ObservationWrapper):
    def __init__(self, env):
        super().__init__(env)
        self.observation_space = spu.update_obs_space(env, {"torso_height": (1,)})

    def observation(self, obs):
        obs = self.obs_temp   ### what I added
        x_pos, y_pos, z_pos = self.sim.data.get_body_xpos("torso/0")
   ........
   ........

After that, I got an error like below.

Process ForkProcess-3:
Traceback (most recent call last):
  File "/home/bak/Projects/metamorph/metamorph/envs/vec_env/subproc_vec_env.py", line 27, in worker
    remote.send([env.reset() for env in envs])
  File "/home/bak/Projects/metamorph/metamorph/envs/vec_env/subproc_vec_env.py", line 27, in <listcomp>
    remote.send([env.reset() for env in envs])
  File "/home/bak/Projects/metamorph/metamorph/envs/wrappers/multi_env_wrapper.py", line 40, in reset
    obs = self._env.reset()
  File "/home/bak/Projects/metamorph/metamorph/algos/ppo/envs.py", line 187, in reset
    observation = super(RecordEpisodeStatistics, self).reset(**kwargs)
  File "/home/bak/anaconda3/envs/metamorph/lib/python3.8/site-packages/gym/core.py", line 323, in reset
    return self.env.reset(**kwargs)
  File "/home/bak/Projects/metamorph/metamorph/algos/ppo/envs.py", line 169, in reset
    return self.env.reset(**kwargs)
  File "/home/bak/anaconda3/envs/metamorph/lib/python3.8/site-packages/gym/wrappers/time_limit.py", line 68, in reset
    return self.env.reset(**kwargs)
  File "/home/bak/anaconda3/envs/metamorph/lib/python3.8/site-packages/gym/wrappers/order_enforcing.py", line 42, in reset
    return self.env.reset(**kwargs)
  File "/home/bak/anaconda3/envs/metamorph/lib/python3.8/site-packages/gym/wrappers/env_checker.py", line 47, in reset
    return self.env.reset(**kwargs)
  File "/home/bak/anaconda3/envs/metamorph/lib/python3.8/site-packages/gym/core.py", line 379, in reset
    obs, info = self.env.reset(**kwargs)
ValueError: too many values to unpack (expected 2)

How can I fix these problems and check the result of your experiments?

I used Anaconda environment with mujoco-2.1.0 rather than Docker due to using free version of mujoco and using mujoco-py. Unfortunately, using mujoco after 2.1.0 does not support mujoco-py due to it has its own python binding.

Code for evaluation and visualization

Hello!

Thanks a lot for the great work and for providing the code!

Are there any code/scripts available for evaluation or visualization?

Thanks a lot your help!

Recommend Projects

React

A declarative, efficient, and flexible JavaScript library for building user interfaces.
Vue.js

🖖 Vue.js is a progressive, incrementally-adoptable JavaScript framework for building UI on the web.
Typescript

TypeScript is a superset of JavaScript that compiles to clean JavaScript output.
TensorFlow

An Open Source Machine Learning Framework for Everyone
Django

The Web framework for perfectionists with deadlines.
Laravel

A PHP framework for web artisans
D3

Bring data to life with SVG, Canvas and HTML. 📊📈🎉

Recommend Topics

javascript

JavaScript (JS) is a lightweight interpreted programming language with first-class functions.
web

Some thing interesting about web. New door for the world.
server

A server is a program made to process requests and deliver data to clients.
Machine learning

Machine learning is a way of modeling and interpreting data that allows a piece of software to respond intelligently.
Visualization

Some thing interesting about visualization, use data art
Game

Some thing interesting about game, make everyone happy.

Recommend Org

Facebook

We are working to build community through open source technology. NB: members must have two-factor auth.
Microsoft

Open source projects and samples from Microsoft.
Google

Google ❤️ Open Source for everyone.
Alibaba

Alibaba Open Source for everyone
D3

Data-Driven Documents codes.
Tencent

China tencent open source team.