How can I use this model with RTX4070Ti and CUDA 11.8? (mmcv version conflict)

[YkY0224]

Hello,

I'm trying to run the model on my RTX4070Ti with a CUDA 11.8 environment. I set up my environment with the following commands:

bash

Install PyTorch, torchvision, torchaudio, and CUDA 11.8 runtime

conda install pytorch==2.1.2 torchvision==0.16.2 torchaudio==2.1.2 pytorch-cuda=11.8 -c pytorch -c nvidia

Install mmcv for CUDA 11.8 and PyTorch 2.1.2

pip install mmcv==2.1.0 -f https://download.openmmlab.com/mmcv/dist/cu118/torch2.1/index.html
However, when I try to run the model, I receive an error that indicates a version incompatibility with mmcv:

mmcv_minimum_version = '1.4.8'
mmcv_maximum_version = '1.6.0'
MMCV==2.1.0 is used but incompatible. Please install mmcv>=1.4.8, <=1.6.0.
It seems that tr3d requires mmcv in the range [1.4.8, 1.6.0], but I installed mmcv 2.1.0 to match CUDA 11.8 and PyTorch 2.1.2.

Could you please advise on how to resolve this conflict?

Should I downgrade mmcv to a version between 1.4.8 and 1.6.0 even though I'm using CUDA 11.8?
Is there a recommended setup to use this model on RTX4070Ti with CUDA 11.8?
Thank you for your help!

[filaPro]

This repo is tested only with package versions provided in our Dockerfile, including cuda 11.3 and torch==1.12.1. I recommend to use only this versions. For later versions you can try tr3d as an official mmdetection3d project, but defining the correct package versions will not be easy.