【MM】CMT：Cross Modal Transformer 环境配置

前言

环境已经调通

mmcv 用pip安装 ；mmdet、mmdet3d、mmseg用源码安装⭐ 一个巨坑：git clone https://github.com/open-mmlab/mmdetection.git对应的是master，也就是最新版本的mmdet，运行pip install -v -e . # or "python setup.py develop"是按最新依赖关系来配置，因此会有因为环境没对齐出现的报错

虚拟环境

# env

conda create -n cmt python=3.8 -y

conda activate cmt

# pytorch 1.9.0+cu111

pip install torch==1.9.0+cu111 torchvision==0.10.0+cu111 torchaudio==0.9.0 -f https://download.pytorch.org/whl/torch_stable.html

MMCV

安装 MMCV — mmcv 1.6.0 文档

pip install mmcv-full==1.6.0 -f https://download.openmmlab.com/mmcv/dist/cu111/torch1.9.0/index.html

mmdetection 源码安装

Prerequisites — MMDetection 2.24.0 documentation 去 github 的 2.24.0 tag 下载 .zip，解压后放到根目录下运行下面代码

# git clone https://github.com/open-mmlab/mmdetection.git

cd mmdetection

pip install -r requirements/build.txt

pip install -v -e . # or "python setup.py develop"

mmdetsegment 源码安装

GitHub - open-mmlab/mmsegmentation at v0.29.1 同上

#git clone https://github.com/open-mmlab/mmsegmentation.git

cd mmsegmentation

pip install -e . # or "python setup.py develop"

MMDectection3D 1.0.0rc5 源码安装

GitHub - open-mmlab/mmdetection3d at v1.0.0rc5 同上

#git clone https://github.com/open-mmlab/mmdetection3d.git

cd mmdetection3d

pip install -v -e . # or "python setup.py develop"

其他

# spconv

pip install spconv-cu111==2.1.21

# flash-attn

pip install flash-attn==0.2.2

# numpy

pip install numpy==1.22.4

eval

CUDA_VISIBLE_DEVICES='0,1' python -m torch.distributed.launch --nproc_per_node=2 --master_port=29505 CMT/tools/train.py --config /root/autodl-tmp/CMT/projects/configs/fusion/cmt_voxel0100_r50_800x320_cbgs.py