段错误 (核心已转储)

[Liu-xiaoyu11]

Hello, @boqian-li ： Your excellent work has given me a great deal of inspiration, and I thank you very much. I have encountered some issues while trying to reproduce the code. Do you have any good suggestions? Regarding whether the prompt in the command-line statement should be enclosed in double quotes: I ran the following commands: CUDA_VISIBLE_DEVICES=1 python launch_garmentdreamer.py --template_path /mnt/sda2/code/GarmentDreamer-main/input_data/ori_meshes/a_American_Western_cowboy_style_jumpsuit_made_of_brown_leather/Jumpsuit_747_translated.obj --prompt a blue denim tee； CUDA_VISIBLE_DEVICES=1 python launch_garmentdreamer.py --template_path /mnt/sda2/code/GarmentDreamer-main/input_data/ori_meshes/a_American_Western_cowboy_style_jumpsuit_made_of_brown_leather/Jumpsuit_747_translated.obj --prompt "a blue denim tee"； CUDA_VISIBLE_DEVICES=1 python launch_garmentdreamer.py --template_path /mnt/sda2/code/GarmentDreamer-main/input_data/ori_meshes/a_American_Western_cowboy_style_jumpsuit_made_of_brown_leather/Jumpsuit_747_translated.obj --prompt "blue denim tee"； All of them resulted in the following issue: I suspect that my prompt settings might be incorrect, but I have tried several variations without success. Do you have any suggestions? I appreciate your help. I look forward to your reply.

[boqian-li]

Hi @Liu-xiaoyu11 , thanks for your interest in our work. It looks like there's a problem with Pytorch Cuda and xformers. Could you please check first with the command python -c "import torch; print(torch.cuda.is_available())"?

And it looks like you are using the xformers which requires cuda==12.1, please also make sure the cuda version for pytorch is aligned with that with the command python -c "import torch; print(torch.version.cuda)"

IF there are any other questions, please let me know.

[Liu-xiaoyu11]

Hi @Liu-xiaoyu11 , thanks for your interest in our work. It looks like there's a problem with Pytorch Cuda and xformers. Could you please check first with the command python -c "import torch; print(torch.cuda.is_available())"?

And it looks like you are using the xformers which requires cuda==12.1, please also make sure the cuda version for pytorch is aligned with that with the command python -c "import torch; print(torch.version.cuda)"

IF there are any other questions, please let me know. Hello, @boqian-li : I just used the following command to try setting CUDA to 12.1,conda install pytorch==2.3.1 torchvision==0.18.1 torchaudio==2.3.1 pytorch-cuda=12.1 -c pytorch -c nvidia. However, this causes issues when I attempt to install diff-gaussian-rasterization. Do you have any suggestions on how to resolve this? I look forward to your reply.

[boqian-li]

hi, actually you don't need to follow the version of cuda that xformers needs, it actually depends on the version of xformers you installed.

Just install cuda runtime with verision==11.8, and make sure it's selected with the command: nvcc --version. Then follow the environment setup instructions in readme.md.

pytorch has its cuda toolkit version (exactly the version in your installation command: like cu113 cu118, etc.) and on your GPU, you all so have different cuda runtime compiler versions (it seems you have CUDA==11.3, you can try to run nvcc --version to find it out)

although it usually works when the two versions are different, I recommend installing the same pytorch-cuda version as the Nvidia cuda compiler version.

[Liu-xiaoyu11]

hi, try this: vllm-project/vllm#1453 Hello, @boqian-li If I match CUDA 11.3 using the following command:pip install torch==1.12.1+cu113 torchvision==0.13.1+cu113 torchaudio==0.12.1 --extra-index-url https://download.pytorch.org/whl/cu113 Will this affect the subsequent running of my code? Thank you very much for your help. I look forward to your reply.

[boqian-li]

I recommend just deleting the virtual env first. check the cuda runtime version with command nvcc --version and make sure you are using the version==11.8 (seems you are using 11.3 at the moment) then follow the setup instructions (pytorch-cuda==11.8 especially) to rebuild the virtual environment from scratch then it should work. Any other questions, let me know please, glad to help!

[Liu-xiaoyu11]

I recommend just deleting the virtual env first. check the cuda runtime version with command nvcc --version and make sure you are using the version==11.8 (seems you are using 11.3 at the moment) then follow the setup instructions (pytorch-cuda==11.8 especially) to rebuild the virtual environment from scratch then it should work. Any other questions, let me know please, glad to help!

Hello, @boqian-li: Thank you very much for your patient responses. The issues mentioned above have been resolved; it was due to incorrect installation commands for xformers, which messed up the environment. Your assistance has been very helpful to me. I have two more small questions: What software did you use to view the final reconstructed .obj file? When I tried viewing it directly in Blender, the texture information did not display. Regarding the items you mentioned on GitHub, will there be further uploads? Once again, thank you very much for your help. I look forward to your reply.