Pytorch environment issues

[yyyzjing]

May I ask what version of pytorch should be installed when cuda version is 11.3?

1、I have tried versions with torch=1.12.1+cu113, torch vision=0.13.1+cu113, and torch studio=0.12.1+cu113. Under these conditions, I will not be able to run the/Deformable DETR/models/ops/test.py file, which will result in an error nvrtc: error: invalid value for -- gpu architecture (- arch).

2、I have tried versions with torch=1.11.0+cu113 torch vision=0.12.0+cu113 torch studio=0.11.0, the/Deformable DETR/models/ops/test.py file can run normally. However, if I run the training main.py, an error will occur as shown in the following figure

i am looking forward to your early reply, thx!!!!!!!

[HELLORPG]

Hi, could you please tell me which GPU type you are using? I think maybe you are using 40xx (like 4090).

[yyyzjing]

yes，i am trying to use 4090 for training

[HELLORPG]

As NVIDIA 4090 is a new GPU type, you should use a newer version of CUDA. I recommend:

conda install pytorch==1.13.1 torchvision==0.14.1 torchaudio==0.13.1 pytorch-cuda=11.7 -c pytorch -c nvidia

Waiting for your good news.

P.S. The second error you've mentioned is an issue caused by the pytorch version. ResNet50_Weights is a newly proposed feature in pytorch.

[yyyzjing]

ok,i will try it again!thx!!!

[HELLORPG]

Oh, and I also suggest you use python=3.10. I remember but not sure that we also use some new features of python in parameter comments.

[yyyzjing]

ok， I also would like to ask, is the version you recommend the lowest version that can run through your code?

发自我的iPhone

------------------ Original ------------------ From: Ruopeng Gao @.> Date: Thu,Nov 30,2023 11:02 PM To: MCG-NJU/MeMOTR @.> Cc: yyyzjing @.>, Author @.> Subject: Re: [MCG-NJU/MeMOTR] Pytorch environment issues (Issue #6)

Oh, and I also suggest you use python=3.10. I remember but not sure that we also use some new features of python in parameter comments.

— Reply to this email directly, view it on GitHub, or unsubscribe. You are receiving this because you authored the thread.Message ID: @.***>

[HELLORPG]

Honestly, I'm not sure about that. I always use the latest version of PyTorch cause I love to catch up with the fashion. So python=3.10 and pytorch=1.13.1 is the only choice during my experiments.

Depending on your exp outputs, I think pytorch 1.12 is also OK. You can also try to use lower pytorch on other GPUs (do not use ResNet50_Weights, but use True like MOTR), but NVIDIA 4090 needs CUDA >= 11.7, which may not suit earlier versions.

For the version of Python, if you get some error about the symbol |, then you should use python=3.10 for your virtual env.

[yyyzjing]

Thank you very much for your reply!!! I will conduct experiments later!  Best wishes!

发自我的iPhone

------------------ Original ------------------ From: Ruopeng Gao @.> Date: Thu,Nov 30,2023 11:19 PM To: MCG-NJU/MeMOTR @.> Cc: yyyzjing @.>, Author @.> Subject: Re: [MCG-NJU/MeMOTR] Pytorch environment issues (Issue #6)

Honestly, I'm not sure about that. I always use the latest version of PyTorch cause I love to catch up with the fashion. So python=3.10 and pytorch=1.13.1 is the only choice during my experiments.

Depending on your exp outputs, I think pytorch 1.12 is also OK. You can also try to use lower pytorch on other GPUs (do not use ResNet50_Weights, but use True like MOTR), but NVIDIA 4090 needs CUDA >= 11.7, which may not suit earlier versions.

For the version of Python, if you get some error about the symbol |, then you should use python=3.10 for your virtual env.

— Reply to this email directly, view it on GitHub, or unsubscribe. You are receiving this because you authored the thread.Message ID: @.***>

[HELLORPG]

You're welcome. Waiting for your good news.

[HELLORPG]

Hey there! Just wanted to check in and see if you have solved the issue. Is there anything else I can do to help with this? Let me know.

[HELLORPG]

As I haven't received your reply for a long time, I am closing this issue temporarily. You can re-open this issue if you need~