Error Running Pour from Cup to Cup task for data collection

[Mi-Przystupa]

I'm using RLbench for generating a dataset and wanted to use the PourFromCuptoCup task for my work. When I run the code it errors out and says this shouldn't happen.

I've included an image of the error and the script that was run (had to upload as txt file)

gen_rl_bench_data.txt

I was running in a conda environment running python 3.9.7 using an Nvidia RTX 3080 graphics card on ubuntu 21.10

[stepjam]

Thanks for raising. How soon after launching the dataset_generation does the error appear?

[Mi-Przystupa]

it varies I think from run to run. Sometimes it is completely fine for pouring into certain positions then when the environment resets to a new setting, it will crash. I think as is the script I sent SHOULD work as is, but I think the most minimal way (hopefully) to reproduce is:

from rlbench.environment import Environment
from rlbench.action_modes import ArmActionMode, ActionMode
from rlbench.observation_config import ObservationConfig
from rlbench.tasks import ReachTarget, PourFromCupToCup

obs_config = ObservationConfig()
obs_config.set_all(True)
action_mode = ActionMode(ArmActionMode.ABS_JOINT_VELOCITY)
env = Environment(
      action_mode, '', obs_config, False)
env.launch()


task = env.get_task(PourFromCupToCup)

for i in range(200):
       demo = task.get_demos(1, live_demos=True)[0]
        

just as a disclaimer, this might not run as is since I was refactoring from the script originally included.