Leo0408
commented
1 year ago We are using NVIDIA's 40 series GPU, and the appropriate corresponding CUDA version is at least 11.7. However, we encounted this ImportError: "libcudart.so.10.1: cannot open shared object file: No such file or directory." , We had to downgrade the CUDA version to 10.1, and then we encountered the error mentioned above. We hope that this detail can be helpful in addressing the issue.
I run FollowingAnything on CUDA 10.1. But when I install spatial_correlation_sampler by using setup.py in the Pytorch-Correlation-extension package it says
File "/home/nickbit_asus/anaconda3/envs/follow_anything/lib/python3.9/site-packages/spatial_correlation_sampler-0.4.0-py3.9-linux-x86_64.egg/spatial_correlation_sampler/__init__.py", line 1, in <module> from .spatial_correlation_sampler import SpatialCorrelationSampler, spatial_correlation_sample File "/home/nickbit_asus/anaconda3/envs/follow_anything/lib/python3.9/site-packages/spatial_correlation_sampler-0.4.0-py3.9-linux-x86_64.egg/spatial_correlation_sampler/spatial_correlation_sampler.py", line 6, in <module> import spatial_correlation_sampler_backend as correlation ImportError: /home/nickbit_asus/anaconda3/envs/follow_anything/lib/python3.9/site-packages/spatial_correlation_sampler-0.4.0-py3.9-linux-x86_64.egg/spatial_correlation_sampler_backend.cpython-39-x86_64-linux-gnu.so: undefined symbol: _ZNK3c106IValue23reportToTensorTypeErrorEvAnd the intall error also happens when I use "pip install", reporting the CUDA version is too low (<11.7), while CUDA version above 11 may generates other CUDA version as well.
Do you have any idea where the problem is and how it can be solved?
By the way, I am very willing to run FollowingAnything on Google Colab. Have you create a Google Colab notebook that can perform object tracking on videos?
Thank you!