Closed SummerLam closed 3 months ago
wanghao9610
commented
3 months ago You need to rebuild the detrex package. You can refer to the following code:
# Suppose you have set root_dir env_var and have installed the requirements on ovdino conda env.
cd $root_dir/ovdino
conda activate ovdino
python -m pip install -e detectron2-717ab9
pip install -e ./If the above processing failed, you may need to recreate a new env following the installation guideline.
SummerLam
commented
3 months ago
i got this error when i run the commands:
python -m pip install -e detectron2-717ab9
pip install -e ./
so this is probably the reason? yet i have recreate a new env but i encountered the same error whenever i tried running these 2 commands. i would like to know if there is any solutions for it?
wanghao9610
commented
3 months ago @SummerLam I need more info to debug your issue, please give me more error information.
wanghao9610
commented
3 months ago It looks like that the gcc version is too slow, you could type the following command:
which gcc; gcc --versionTry to install gcc by conda:
conda install gcc=9 gxx=9 -c conda-forge -y # Optional: install gcc9Then recompile the detecton2 and detrex package.
SummerLam
commented
3 months ago
Requirement already satisfied: portalocker in /opt/conda/envs/ovdino/lib/python3.10/site-packages (from iopath<0.1.10,>=0.1.7->detectron2==0.6) (2.10.1)
this is the captured error message after running the command python -m pip install -e detectron2-717ab9
wanghao9610
commented
3 months ago The error is your cuda not install correctly, you can get help from the issue.
SummerLam
commented
3 months ago ok, thank you so much for the information provided and your help. i will read that through.
SummerLam
commented
3 months ago
Hi, i have followed the steps here https://github.com/wanghao9610/OV-DINO/issues/14#issuecomment-2277355677 tried specifying the cuda_home path again and also tried running the command conda install pytorch==1.13.1 torchvision==0.14.1 torchaudio==0.13.1 pytorch-cuda=11.6 -c pytorch -c nvidia -y. However, this issue still exists.
wanghao9610
commented
3 months ago The figure shows that your installed cuda version is 12.1, while your torch cuda version is 11.6, the two cudas are not match well. You need to install cuda 11.6 manually or install higher version pytorch with cuda12.1(not recommend, may introduce more issue).
SummerLam
commented
3 months ago I see, thank you so much for your kind assistance.
SummerLam
commented
3 months ago Hi, i think i am facing this same issue again even though i have had cuda11.6 installed, what else did i miss?
wanghao9610
commented
3 months ago Do you have compiled detectron2 and detrex successfully? The error info shows that your detrex is not build successfully, as the deformable_conv operator is not installed.
Cloud65000
commented
3 months ago I encountered this problem after run " python3 setup.py build --force develop" ![Uploading Problem.jpg…]()
Cloud65000
commented
3 months ago I encountered this problem after run " python3 setup.py build --force develop" ![Uploading Problem.jpg…]()
Cloud65000
commented
3 months ago The content of the error is shown like this :
on2/layers/csrc/vision.cpp:111
Exception raised from registerLibrary at ../aten/src/ATen/core/dispatch/Dispatcher.cpp:180 (most recent call first):
frame #0: c10::Error::Error(c10::SourceLocation, std::string) + 0x57 (0x7f50e0223617 in /usr/local/python3.9.18/lib/python3.9/site-packages/torch/lib/libc10.so)
frame #1: c10::detail::torchCheckFail(char const, char const, unsigned int, std::string const&) + 0x64 (0x7f50e01de98d in /usr/local/python3.9.18/lib/python3.9/site-packages/torch/lib/libc10.so)
frame #2: c10::Dispatcher::registerLibrary(std::string, std::string) + 0x3c5 (0x7f507944cb05 in /usr/local/python3.9.18/lib/python3.9/site-packages/torch/lib/libtorch_cpu.so)
frame #3: torch::Library::Library(torch::Library::Kind, std::string, c10::optional
wanghao9610
commented
3 months ago 1) I can't see the picture you uploaded. 2) The error looks like you didn't activate enviroment well, or you don't use the conda? If you haven't installed anaconda, you may install anaconda first... 3) The most possiable reson is pytorch installation error, you need first to test the pytorch installation:
python -c "import torch; print(torch.__version__); print(torch.cuda.is_available())"
- I can't see the picture you uploaded.
- The error looks like you didn't activate enviroment well, or you don't use the conda? If you haven't installed anaconda, you may install anaconda first...
- The most possiable reson is pytorch installation error, you need first to test the pytorch installation:
python -c "import torch; print(torch.__version__); print(torch.cuda.is_available())"
the issue has been fixed, after specifying the cuda path to be cuda-11.6, we have to run the command pip install -e ./ again and then it can work
SummerLam
commented
3 months ago Thank you for your help and attention
Cloud65000
commented
3 months ago I didn't use conda.
- I can't see the picture you uploaded.
- The error looks like you didn't activate enviroment well, or you don't use the conda? If you haven't installed anaconda, you may install anaconda first...
- The most possiable reson is pytorch installation error, you need first to test the pytorch installation:
python -c "import torch; print(torch.__version__); print(torch.cuda.is_available())"
I didn't use conda. I'm in the docker container with cuda2.1 and torch12.1. I'll try to use conda and strictly follow the README and try again.
when i tried to run the sh scripts/demo.sh process, i got this error of failing to import detrex._C and i followed the instruction online of how to fix this issue which is by setting the cuda_home path, tho my cuda_home is not none, i ran the command (export CUDA_HOME=/usr/local/cuda) anyway. However, it didnt resolve the issue. therefore, i would like to seek help here to see what else can be done to fix the error.