Plans for releasing Docker

[kjeiun]

Do you have any plans for releasing docker?

• ) Do you know how many time it will take to generate all demonstrations?

I'm using 1 L40S gpu and there is no more progress than the shown Images (Almost 1 hr has passed).. Does it take longer than 1hr for close jar demo? Or there is some problem with generating the demonstrations?

There's no output in the train_data folder.

[GuanxingLu]

The data generation process should be normal like this:

The process requires no GPU resource.
We are working on releasing Dockerfile but have encountered some issues, stay tuned!

[kjeiun]

Sounds great :)
Actually I found that the problem was because of using RLBench inside of the docker.
Are you having the same problem as

Or do you have docker file at least for RLBench?

[GuanxingLu]

Have you tried xvfb in a headless machine?

Sorry, I have no Dockerfile for RLBench. Besides, maybe you can refer to stepjam/RLBench#176 to use docker image with coppeliasim and pyrep.

[kjeiun]

Yes I did. I tried to use that docker but the docker was not installed with cuda , so i tried to install it by myself but it keeps failing.

So I think it would be better to wait you releasing dockerfile.

Thank you :)

[GuanxingLu]

Ok, i see, have you tried everything in a physical server without docker?

[kjeiun]

I'm a using a lab server so I only need to work with docker :(

[kjeiun]

I can generate the demonstration from the docker you mentioned, but since the cuda is not installed from that docker, I think I can just generate the demonstrations from there.

is it possible to train manigaussian without installing RLBench? Because I can get the data from that docker and move it to other cuda-available docker just for training ( I don't need env.launch() while "training" process right?)

[GuanxingLu]

Yeh, you can train manigaussian without RLBench env launch (but may use some tools from RLBench, e.g., the script that convert rgbd to point cloud, I think they can work properly in docker container).

[kjeiun]

Sounds great! Thanks for helping me. I'll leave the comment for others if everything goes well.

[GuanxingLu]

Okay, just keep this issue. If you encounter any other problems on training manigaussian in docker, feel free to ask!

[kjeiun]

I also got same error with mask2former :( They say there's a circular import)

[GuanxingLu]

I also got same error with mask2former :( They say there's a circular import)

Let's keep the discussion in #11

[kjeiun]

I found that the RLBench problem when using docker is solved when installing Coppeliasim_Edu_v_4_1_0 rather than any other version.