How to install InternImage/detection with A100

[Deokhwan-Kyeong]

I already installed

python 3.9
pytorch 2.0.1
torchvision 0.15.2
torchaudio 2.0.2
cuda 11.8

and I command
pip install -U openmim
mim install mmcv-full==1.5.0

then, I faced this situation
Installing mmcv-full is taking too long.

How can I solve this problem?

(intern) vv@bninc-DGX-Station-A100:~/aaa/InternImage/detection$ mim install mmcv-full==1.5.0
Looking in links: https://download.openmmlab.com/mmcv/dist/cu118/torch2.0.0/index.html
Collecting mmcv-full==1.5.0
Downloading mmcv-full-1.5.0.tar.gz (530 kB)
━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━ 530.8/530.8 kB 27.0 MB/s eta 0:00:00
Preparing metadata (setup.py) ... done
Collecting addict (from mmcv-full==1.5.0)
Downloading addict-2.4.0-py3-none-any.whl (3.8 kB)
Requirement already satisfied: numpy in /home/vv/anaconda3/envs/intern/lib/python3.9/site-packages (from mmcv-full==1.5.0) (1.26.2)
Collecting packaging (from mmcv-full==1.5.0)
Downloading packaging-23.2-py3-none-any.whl.metadata (3.2 kB)
Requirement already satisfied: Pillow in /home/vv/anaconda3/envs/intern/lib/python3.9/site-packages (from mmcv-full==1.5.0) (10.0.1)
Requirement already satisfied: pyyaml in /home/vv/anaconda3/envs/intern/lib/python3.9/site-packages (from mmcv-full==1.5.0) (6.0.1)
Collecting yapf (from mmcv-full==1.5.0)
Downloading yapf-0.40.2-py3-none-any.whl.metadata (45 kB)
━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━ 45.4/45.4 kB 23.9 MB/s eta 0:00:00
Collecting opencv-python>=3 (from mmcv-full==1.5.0)
Downloading opencv_python-4.8.1.78-cp37-abi3-manylinux_2_17_x86_64.manylinux2014_x86_64.whl.metadata (19 kB)
Requirement already satisfied: importlib-metadata>=6.6.0 in /home/vv/anaconda3/envs/intern/lib/python3.9/site-packages (from yapf->mmcv-full==1.5.0) (7.0.0)
Collecting platformdirs>=3.5.1 (from yapf->mmcv-full==1.5.0)
Downloading platformdirs-4.1.0-py3-none-any.whl.metadata (11 kB)
Collecting tomli>=2.0.1 (from yapf->mmcv-full==1.5.0)
Downloading tomli-2.0.1-py3-none-any.whl (12 kB)
Requirement already satisfied: zipp>=0.5 in /home/vv/anaconda3/envs/intern/lib/python3.9/site-packages (from importlib-metadata>=6.6.0->yapf->mmcv-full==1.5.0) (3.17.0)
Downloading opencv_python-4.8.1.78-cp37-abi3-manylinux_2_17_x86_64.manylinux2014_x86_64.whl (61.7 MB)
━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━ 61.7/61.7 MB 36.0 MB/s eta 0:00:00
Downloading packaging-23.2-py3-none-any.whl (53 kB)
━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━ 53.0/53.0 kB 3.8 MB/s eta 0:00:00
Downloading yapf-0.40.2-py3-none-any.whl (254 kB)
━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━ 254.7/254.7 kB 75.0 MB/s eta 0:00:00
Downloading platformdirs-4.1.0-py3-none-any.whl (17 kB)
Building wheels for collected packages: mmcv-full
Building wheel for mmcv-full (setup.py) ... /

[IndefiniteBen]

How long does it take and what CPU are you using?

When I run it on Colab (free with a T4 GPU) it takes about 30 minutes to build the wheel for mmcv-full.

I think if you instead set it up with Python 3.7 it doesn't need to build the wheel and is therefore much faster, but if the wheel needs to be built, it takes a long time.