application hindsight

[KuangSongyang1113]

If I want to apply hindsight to my project, do I need to prepare my dataset just like lyft dataset?

[YurongYou]

Hi KuangSongyang1113! Thanks for your interest! Yes, you can prepare your dataset as described in the repo.

[KuangSongyang1113]

By installing spconv with python setup.py bdist_wheel，I got "Caffe2: CUDA detected:10.1. But actually I got cuda version 11.8 with nvidia-smi command in terminal. How did spconv check the cuda version?

[YurongYou]

I am not exactly sure, but it might be also related to the cudatoolkit you installed via conda. Did it throw an error? If not it should be fine, since cuda should be backward compatible.