search

mees / calvin

CALVIN - A benchmark for Language-Conditioned Policy Learning for Long-Horizon Robot Manipulation Tasks
http://calvin.cs.uni-freiburg.de
MIT License
366 stars 55 forks source link

TypeError: Error instantiating 'calvin_env.envs.play_table_env.PlayTableSimEnv' : expected str, bytes or os.PathLike object, not NoneType disconnecting id 0 from server #69

Open LPY1219 opened 7 months ago

LPY1219 commented 7 months ago

Hi, Thanks for your efforts on building such an amazing benchmark! I am trying to use it to validate my RL algorithm, but i met some problems when I tried the RL jupyter you provide in the repository. I can successfully run the script by using google colab, but when i transform it into .py file and install the related packages, it can not work. I can successfuly load the package by " from calvin_env.envs.play_table_env import PlayTableSimEnv". I do not know why the program will give me the above error when it begins to execute " env = hydra.utils.instantiate(cfg.env) ". Can you help me ?

lukashermann commented 7 months ago

Hi, @LPY1219, I did what you described and tried to extract the code from the colab to a .py file:

import hydra
import numpy as np
import cv2
from hydra import initialize, compose

with initialize(config_path="./calvin_env/conf"):
  cfg = compose(config_name="config_data_collection.yaml", overrides=["cameras=static_and_gripper"])
  cfg.env["use_egl"] = False
  cfg.env["show_gui"] = False
  cfg.env["use_vr"] = False
  cfg.env["use_scene_info"] = True
  print(cfg.env)

env = hydra.utils.instantiate(cfg.env)
observation = env.reset()
#The observation is given as a dictionary with different values
print(observation.keys())
for i in range(5):
  # The action consists in a pose displacement (position and orientation)
  action_displacement = np.random.uniform(low=-1, high=1, size=6)
  # And a binary gripper action, -1 for closing and 1 for oppening
  action_gripper = np.random.choice([-1, 1], size=1)
  action = np.concatenate((action_displacement, action_gripper), axis=-1)
  observation, reward, done, info = env.step(action)
  rgb = env.render(mode="rgb_array")[:,:,::-1]
  cv2.imshow("win", rgb)
  cv2.waitKey(1)

It runs without errors on my computer, can you try it yourself?

LPY1219 commented 7 months ago

Thanks for your reply! I have sovled it by adding init file in the "calvin_env" directory. And then the program can successfully find the package.

lukashermann commented 7 months ago

@LPY1219 Great that you solved your issue! Just out of curiosity, did you follow the installation instructions in our README? calvin_env should get installed with pip in the install.sh script, so you wouldn't neet an init file.

initial-h commented 1 month ago

@LPY1219 Great that you solved your issue! Just out of curiosity, did you follow the installation instructions in our README? calvin_env should get installed with pip in the install.sh script, so you wouldn't neet an init file.

I installed following README, but still need init.py to solve the same error.

FutureTwT commented 3 weeks ago

Perhaps because the calvin_env library is cloned in the current directory, the import process looked for the local path instead of the installation path