gregor

Hi @sadeghkt . Sure, I'm currently working on making sure the docker image we provide still works. You would have to install the [nvidia-container-toolkit](https://docs.nvidia.com/datacenter/cloud-native/container-toolkit/latest/install-guide.html) on your end. This will allow...

Hi @aopolin-lv, sorry for the delayed response. Did you solve the issue on the RLBench side?. I noticed that you also posted a similar issue stepjam/RLBench#240, and it seems it's...

Hi @ubless607 . That could be caused by the number of cameras being active, although last time I checked it didn't run that slow. I'll check on my end running...

Hi @ubless607 , sorry for the delay. I replied to a similar issue here #7 . Basically, could you test if things improve when using fewer cameras ?. Were you...

Hi @david-dorf . I think I ran into a similar issue the other day when using wayland. Could you check which version of `glfw` are you using? ```bash pip freeze...

Hi @Zed-Wu . Sorry, this is an issue on our docs. We added some warnings on which version of RLBench to install. Most likely you're using the [upstream](https://github.com/stepjam/rlbench) version (the...

Sorry for the delay. Uhmmmm, there are a few classes that use this `ignore_collision` field. What I usually do is to create a separate virtual environment, either with conda or...

Hi @kfc2333 , sorry for the late reply. I checked the RLBench codebase and it seems that this image contains an object ID per pixel. It's like when doing semantic...

Also, I think you can use [this](https://github.com/stepjam/RLBench/issues/72) to convert your IDs to an object name.

Hi @Jianshu-Hu , sorry for the delayed response. I ran the `visualize_task` script and indeed, the task fails to collect demonstrations because of not being able to reset properly. Here's...