jinlinyi
Testing on RTX3090
Dear respected authors,
let me first congrats you all on the great work!
I am trying to run your method On Linux Ubuntu 18.04, but I am facing the following problem:
/home/desk/anaconda3/envs/sparseplane/lib/python3.7/site-packages/torch/cuda/init.py:125: UserWarning: NVIDIA GeForce RTX 3090 Ti with CUDA capability sm_86 is not compatible with the current PyTorch installation. The current PyTorch install supports CUDA capabilities sm_37 sm_50 sm_60 sm_61 sm_70 sm_75 compute_37. If you want to use the NVIDIA GeForce RTX 3090 Ti GPU with PyTorch, please check the instructions at https://pytorch.org/get-started/locally/
warnings.warn(incompatible_device_warn.format(device_name, capability, " ".join(arch_list), device_name))
I did some search and got to know that my RTX 3090 needs sm_86, which Cuda 10.1 doesn't support. Therefore, I removed the virtual env and installed it again, this time with cuda 11.1. I tried installing it using Pip & Cuda as follows:
pip install torch==1.10.1+cu111 torchvision==0.11.2+cu111 torchaudio==0.10.1 -f https://download.pytorch.org/whl/cu111/torch_stable.html
conda install pytorch==1.8.0 torchvision==0.9.0 torchaudio==0.8.0 cudatoolkit=11.1 -c pytorch -c conda-forge
and chose the right detectron2 version from:
https://detectron2.readthedocs.io/en/latest/tutorials/install.html#install-pre-built-detectron2-linux-only
which is:
for torch1.1 (for pip option) :
python -m pip install detectron2 -f
https://dl.fbaipublicfiles.com/detectron2/wheels/cu111/torch1.10/index.html
for torch1.8 (for conda option):
python -m pip install detectron2 -f
https://dl.fbaipublicfiles.com/detectron2/wheels/cu111/torch1.8/index.html
When I tested with the Conda option, I got the following error:
Traceback (most recent call last):
File "tools/inference_sparse_plane.py", line 15, in
When I tested with Pip option, I got the following error:
Traceback (most recent call last):
File "tools/inference_sparse_plane.py", line 15, in
After some research, I got to know that these errors have to do with cuda10.1, which I can't use for the reasons mentioned earlier.
Do you have any recommendations or directions to follow to solve this problem?
Thank you in advance for your guidance.
