nickgkan / 3d_diffuser_actor

Code for the paper "3D Diffuser Actor: Policy Diffusion with 3D Scene Representations"
https://3d-diffuser-actor.github.io/
MIT License
159 stars 16 forks source link

When trying to evaluate on the RLBench test dataset, an error occurs. #34

Closed liops-seungyeob closed 1 month ago

liops-seungyeob commented 1 month ago

When trying to evaluate on the RLBench test dataset, an error occurs.

The log is as follows:

Error: signal 11:

/home/sy/code/3d_diffuse/3d_diffuser_actor/PyRep/CoppeliaSim_Edu_V4_1_0_Ubuntu20_04/libcoppeliaSim.so.1(_Z11_segHandleri+0x30)[0x7f04882b1ae0] /lib/x86_64-linux-gnu/libc.so.6(+0x43090)[0x7f059a2ba090] /home/sy/code/3d_diffuse/3d_diffuser_actor/PyRep/CoppeliaSim_Edu_V4_1_0_Ubuntu20_04/libsimExtDynamicsBullet-2-78.so(_Z19getSimProcAddressesPv+0x42)[0x7f03e45059b2] /home/sy/code/3d_diffuse/3d_diffuser_actor/PyRep/CoppeliaSim_Edu_V4_1_0_Ubuntu20_04/libsimExtDynamicsBullet-2-78.so(simStart+0x8e)[0x7f03e452147e] /home/sy/code/3d_diffuse/3d_diffuser_actor/PyRep/CoppeliaSim_Edu_V4_1_0_Ubuntu20_04/libcoppeliaSim.so.1(_ZN7CPlugin4loadEv+0xa7)[0x7f0488479e17] /home/sy/code/3d_diffuse/3d_diffuser_actor/PyRep/CoppeliaSim_Edu_V4_1_0_Ubuntu20_04/libcoppeliaSim.so.1(ZN16CPluginContainer9addPluginEPKcS1+0x3df)[0x7f0488481eaf] /home/sy/code/3d_diffuse/3d_diffuser_actor/PyRep/CoppeliaSim_Edu_V4_1_0_Ubuntu20_04/libcoppeliaSim.so.1(_ZN9CUiThread27_executeCommandViaUiThreadEP16SUIThreadCommandS1+0x247a)[0x7f048841179a] /home/sy/code/3d_diffuse/3d_diffuser_actor/PyRep/CoppeliaSim_Edu_V4_1_0_Ubuntu20_04/libcoppeliaSim.so.1(ZN9CUiThread25executeCommandViaUiThreadEP16SUIThreadCommandS1+0x3d)[0x7f048841219d] /home/sy/code/3d_diffuse/3d_diffuser_actor/PyRep/CoppeliaSim_Edu_V4_1_0_Ubuntu20_04/libcoppeliaSim.so.1(_Z22simLoadModuleinternalPKcS0+0x3fc)[0x7f04882b572c] /home/sy/code/3d_diffuse/3d_diffuser_actor/PyRep/CoppeliaSim_Edu_V4_1_0_Ubuntu20_04/libcoppeliaSim.so.1(_Z13simulatorInitv+0xb2a)[0x7f04882d43da]

The error occurs at the following part of the code in utils_with_rlbench.py:

def evaluate_task_on_multiple_variations( self, task_str: str, max_steps: int, num_variations: int, # -1 means all variations num_demos: int, actioner: Actioner, max_tries: int = 1, verbose: bool = False, dense_interpolation=False, interpolation_length=100, num_history=1, ): self.env.launch() # Error occurs here

The examples provided by PyRep and RLBench work fine. The same error occurs even when the headless setting is True, indicating a potential issue with the dataset.

However, my dataset follows the structure as described here:

DATA_PREPARATION_RLBENCH.md

What could be the problem?

twke18 commented 1 month ago

I personally didn't see this error message. But I doubt if it's the problem of the dataset. Probably you could try to launch RLBench with this script?

Meanwhile, the command for installing CoppeliaSim assumes a operation system of Ubuntu 20.04. If you are using older distribution, you could modify the file path from https://downloads.coppeliarobotics.com/V4_1_0/CoppeliaSim_Edu_V4_1_0_Ubuntu20_04.tar.xz to https://downloads.coppeliarobotics.com/V4_1_0/CoppeliaSim_Edu_V4_1_0_Ubuntu18_04.tar.xz, and re-install the Coppelia simulator.

liops-seungyeob commented 1 month ago

@twke18

I am using Ubuntu 20.04, so the simulator version is not the issue.

The problem occurs when running eval_peract.sh, specifically during the process of launching RLBench/rlbench/task_design.ttt with PyRep.

However, when running the example code, there is no issue.

The example code is executed as follows: pr = PyRep() pr.launch("task_design.ttt", blocking=False, headless=False) pr.start() But when I run the exact same code in online_evaluation_rlbench/evaluate_policy.py, the aforementioned issue occurs.

I believe the problem arises when running threads in PyQT, but I have not been able to find a solution.

Let me know if you need any further assistance!

liops-seungyeob commented 1 month ago

Problem solved. Just be careful not to mess up anaconda's environment variables.

twke18 commented 1 month ago

Great!

haotian0702 commented 1 month ago

I met the same error. Is it possible for you to share the solution in details? @liops-seungyeob Many thanks!

rjgpinel commented 4 weeks ago

Hi @haotian0702,

Importing pyrep as the first line in the program worked for me.

I just added:

import pyrep

as the first line of evaluate_policy.py