faiss
faiss copied to clipboard
faiss-gpu 1.7.4 install failure
Summary
I follow the official instruction to install faiss-gpu along with pytorch and cuda. I got this error when I run my program:
UserWarning: CUDA initialization: The NVIDIA driver on your system is too old (found version 11040). Please update your GPU driver by downloading and installing a new version from the URL: http://www.nvidia.com/Download/index.aspx Alternatively, go to: https://pytorch.org/ to install a PyTorch version that has been compiled with your version of the CUDA driver. (Triggered internally at /opt/conda/conda-bld/pytorch_1699449200967/work/c10/cuda/CUDAFunctions.cpp:108.)
return torch._C._cuda_getDeviceCount() > 0
command nvidia-smi
shows:
NVIDIA-SMI 470.57.02 Driver Version: 470.57.02 CUDA Version: 11.4
In my conda env, there is no cuda 11040 at all:
cuda-cudart 12.1.105 0 nvidia
cuda-cupti 12.1.105 0 nvidia
cuda-libraries 12.1.0 0 nvidia
cuda-nvrtc 12.1.105 0 nvidia
cuda-nvtx 12.1.105 0 nvidia
cuda-opencl 12.3.101 0 nvidia
cuda-runtime 12.1.0 0 nvidia
cudatoolkit 11.4.1 h8ab8bb3_9 nvidia
pytorch 2.1.1 py3.9_cuda12.1_cudnn8.9.2_0 pytorch
pytorch-cuda 12.1 ha16c6d3_5 pytorch
Platform
OS: Debian 10
Faiss version: 1.7.4
Installed from: conda install faiss-gpu=1.7.4 mkl=2021 pytorch pytorch-cuda numpy -c pytorch -c nvidia
Faiss compilation options: see above
Running on: GPU v100 (same error on A100)
Interface: Python 3.10
Reproduction instructions
create a conda python 3.10 env and install faiss-gpu 1.7.4 by command:
conda install faiss-gpu=1.7.4 mkl=2021 pytorch pytorch-cuda numpy -c pytorch -c nvidia
I suggest forcing pytorch-cuda and cudatoolkit to 11.8:
conda install faiss-gpu=1.7.4 mkl=2021 pytorch pytorch-cuda=11.8 cudatoolkit=11.8 numpy -c pytorch -c nvidia
Let me know if this works.