search

allenai / procthor-10k

The ProcTHOR-10K Houses Dataset
https://procthor.allenai.org/
Apache License 2.0
85 stars 6 forks source link

error about vulkaninfo #11

Open liuxz-cs opened 1 year ago

liuxz-cs commented 1 year ago

I use ai2thor==5.0.0 to load the procthor-10k dataset， but get the following error ?

RuntimeError: vulkaninfo failed to run, please ask your administrator to install vulkaninfo (e.g. on Ubuntu systems this require s running sudo apt install vulkan-tools).

I used the older version of procthor-10k before, and it did not have error, so how can I solve this problem?

liuxz-cs commented 1 year ago

And how to use the old version prior.load_dataset("procthor-10k", revision="ab3cacd0fc17754d4c080a3fd50b18395fae8647"), I use the ai2thor version of 0+391b3fae4d4cc026f1522e5acf60953560235971 and use the code prior.load_dataset("procthor-10k", revision="ab3cacd0fc17754d4c080a3fd50b18395fae8647"), but I cannot get the return results of the event. For example, when I use event = controller.step(action = "GetReachablePositions") nav_pts = event.metadata["actionReturn"], the return is None.

raminnakhli commented 1 year ago

I'm getting the same error on the colab version of the code. Is there any solution for this?

liuxz-cs commented 1 year ago

I'm getting the same error on the colab version of the code. Is there any solution for this?

I have used the normal version of the code, I am not sure about the colab version. But I think you could check the codes, if there exists some operation of cuda, for example .cuda(). This error comes from the line 1048 of controller.py of ai2thor in https://github.com/allenai/ai2thor/blob/main/ai2thor/controller.py. If you use the headless or colab version and use cuda at the same, the program would go into this branch, and you may need to install this vulkan-tool.

Actually, I do not use the headless version, but I meet the same error. I think this code in line 1048 may exist some problem. My machine have gpu and cuda, and I use .cuda() to put my model to the gpu. But my program would also go into this branch, because self.gpu_device is not None. I think this code in line 1048 should be that if the version is headless and self.gpu_device is not None, then go into this branch. So maybe you can modify this code around line 1048 to get the effect that you want in different version.

Hope this can be helpfu for you.