No module dimp_alpha_065
fedorzh opened this issue · comments
After following https://github.com/MasterBin-IIAU/AlphaRefine/tree/master/AlphaRefine_submit, I am getting this in the logs:
ModuleNotFoundError: No module named 'dimp_alpha_065'
@fedorzh
Hi, the possible reason is that you should modify
Ok, you are right, thank you. Now I am getting ModuleNotFoundError: No module named 'pytracking'
though.
And if I hack around that one, the following error is THCudaCheck FAIL file=/opt/conda/conda-bld/pytorch_1556653114079/work/aten/src/THC/THCGeneral.cpp line=51 error=38 : no CUDA-capable device is detected
I think the package is missing a right setup.py file
Hi, ModuleNotFoundError: No module named 'pytracking'
may be caused by the environment variable problem.
Have you added the real path of AlphaRefine_submit to your bashrc file? (as mentioned in README under AlphaRefine_submit) export PYTHONPATH=<path_of_AlphaRefine_submit>:$PYTHONPATH
After adding new environment variables, please run source ~/.bashrc
to make the new variables effective.
Then you should open a new terminal to run this project. It will load your newest environment variables.
And if I hack around that one, the following error is
THCudaCheck FAIL file=/opt/conda/conda-bld/pytorch_1556653114079/work/aten/src/THC/THCGeneral.cpp line=51 error=38 : no CUDA-capable device is detected
I think the package is missing a right setup.py file
Hi, could you tell me your detailed hardware and software types? (such as GPU types, CUDA version, and Pytorch version, etc)
Did you install the pytorch using our command provided in README? conda install pytorch==1.2.0 torchvision==0.4.0 cudatoolkit=10.0 -c pytorch
Python 3.7.9
Cuda compilation tools, release 10.0, V10.0.130
1 GPU Tesla V100
Pytorch was installed beforehand, as I need to test another tracker as well, version is 1.1.0
, torchvision is 0.3.0
Python 3.7.9
Cuda compilation tools, release 10.0, V10.0.130
1 GPUTesla V100
Pytorch was installed beforehand, as I need to test another tracker as well, version is1.1.0
, torchvision is0.3.0
Hi, I think there might be the following possible reasons for this problem.
- Your driver for GPU is broken. Please run
nvidia-smi
to check it. - There exist some conflict between your system's CUDA version and your pytorch version. I strongly suggest to temporally uninstall pytorch and torchvision in your current environment, then install pytorch1.1.0 using the following command.
conda install pytorch==1.1.0 torchvision==0.3.0 cudatoolkit=10.0 -c pytorch
Well, I run Ocean tracker without problems on this driver and the same Anaconda environment.
nvidia-smi
shows all good.
Also, I was able to run this AlphaRef tracker from an external python file, not through exp.sh
, all works fine. But that doesn't run the VOT2020 evaluation
Ok, is the problem still be no CUDA-capable device is detected
now?
I think if AlphaRefine can be run normally without vot-toolkit, CUDA-capable device is detected
problem should not exist now.
no the problem still exists if i run the toolkit way but does not exist if i run directly
It is weird. I don't figure out why it happens. I'm sorry.
But since our code can be run normally without using vot-toolkit, it seems that the problem is not about AlphaRefine itself.
So I guess this issue could be closed because the problem with AlphaRefine has been solved.