【具身智能】Habitat挑战赛2023环境搭建

Habitat挑战赛2023环境搭建

安装步骤

拉取源码

git clone https://github.com/facebookresearch/habitat-challenge.git
cd habitat-challenge

Nvidia容器Tookit

curl -fsSL https://nvidia.github.io/libnvidia-container/gpgkey | sudo gpg --dearmor -o /usr/share/keyrings/nvidia-container-toolkit-keyring.gpg \
  && curl -s -L https://nvidia.github.io/libnvidia-container/stable/deb/nvidia-container-toolkit.list | \
    sed 's#deb https://#deb [signed-by=/usr/share/keyrings/nvidia-container-toolkit-keyring.gpg] https://#g' | \
    sudo tee /etc/apt/sources.list.d/nvidia-container-toolkit.list

sed -i -e '/experimental/ s/^#//g' /etc/apt/sources.list.d/nvidia-container-toolkit.list

sudo apt-get update
sudo apt-get install -y nvidia-container-toolkit

配置docker

这条命令修改了/etc/docker/daemon.json，以保证docker可以在运行时使用NVIDIA的Container

sudo nvidia-ctk runtime configure --runtime=docker

重启docker

sudo systemctl restart docker

nvidia-ctk runtime configure --runtime=docker --config=$HOME/.config/docker/daemon.json

重启docker

1. systemd方式

sudo systemctl --user restart docker

2. snap方式

sudo snap restart docker

验证结果

sudo docker run --rm --runtime=nvidia --gpus all ubuntu nvidia-smi

切换到项目的docker目录

cd /home/moresweet/gitCloneZone/habitat-challenge

在你需要的DockerFile中添加RUN /bin/bash -c ". activate habitat;"
（例如vim ObjectNav_random_baseline.Dockerfile）

docker build . --file docker/ObjectNav_random_baseline.Dockerfile -t objectnav_submission

拉取完成
下载数据集
https://matterport.com/partners/facebook
https://github.com/matterport/habitat-matterport-3dresearch?tab=readme-ov-file

创建链接，将数据集链接到挑战赛的目录中

cd /home/moresweet/gitCloneZone/habitat-challenge
mkdir -p habitat-challenge-data/data/scene_datasets
# 使用训练集创建链接
ln -f -s /media/moresweet/File/hm3d/hm3d-minival-habitat-v0.2 habitat-challenge-data/data/scene_datasets/hm3d_v0.2

修改脚本vim scripts/test_local_objectnav.sh

-v $(realpath habitat-challenge-data/data/scene_datasets/hm3d_v0.2):/habitat-challenge-data/data/scene_datasets/hm3d_v0.2 \

此时，验证集等划分按照https://arxiv.org/pdf/2211.15876.pdf

Krantz J, Lee S, Malik J, et al. Instance-Specific Image Goal Navigation: Training Embodied Agents to Find Object Instances[J]. arXiv preprint arXiv:2211.15876, 2022.

我们暂时模拟一下，在mini数据集的目录中创建链接，以适应程序中对于minival的查找

cd /media/moresweet/File/hm3d/hm3d-minival-habitat-v0.2
ln -f -s . ./minival

验证：

./scripts/test_local_objectnav.sh --docker-name objectnav_submission

验证成功