Scene Graph Generation Note 1

本文记录一下Hiwi的相关笔记。

how to make

ssh ssh1

ssh tntpc

ssh persephone

source ~/.git-prompt.sh

source ~/miniconda3/bin/activate
 :$PATH

export LANG=en_US.UTF-8

neural-motifs

env
1. conda
2. conda install python=3.6
3. conda install pytorch=0.3.0 torchvision=0.2.0 cuda90 -c pytorch
4. conda install cython
5. conda install h5py
6. pycocotools:
  git clone https://github.com/pdollar/coco.git | make | python setup.py build_ext install
7. conda install dill
8. conda install pandas
9. pip install overrides
dataset:
/data/scene_understanding/Visual_Genome/
pycharm:
sh ~/tmp/pycharm-community-2020.2.1/bin/pycharm.sh &

1	export PYTHONPATH=/home/xujing/Dokumente/neural-motifs/:/home/xujing/Dokumente/neural-motifs/lib/

DYNSG

env

conda install pytorch=1.0.0 torchvision=0.2.1 cuda100 -c pytorch
opencv
conda install -c https://conda.anaconda.org/menpo opencv
pip install opencv-python
pip install easydict

export PATH=/usr/local/cuda-10.0/bin:$PATH
export PYTHONPATH=/home/xujing/Dokumente/DynSG/fasterrcnn/lib/:/home/xujing/Dokumente/DynSG/
export PYTHONPATH=$PYTHONPATH:/home/xujing/miniconda3/envs/dynsg/lib/
export LD_LIBRARY_PATH=/home/xujing/miniconda3/envs/dynsg/lib/

issues

_C.cpython-36m-x86_64-linux-gnu.so
pip install -U torchvision==0.4.0
https://ucasfl.github.io/2017/12/21/fix-the-problem-of-that-usr-lib-x86-64-linux-gnu-libstdc-so-6-version-GLIBCXX-3-4-22-not-found/

如何更换 GPU
需要重新编译fasterrcnn

1 2	cd fasterrcnn/lib python setup.py build develop

cuda

import torch

torch.cuda.is_available()
print(torch.version.cuda)
torch.backends.cudnn.enabled

python -c “import torch;print(torch.version.cuda)”

CUDA_VISIBLE_DEVICES=3

nvidia-smi

conda

source activate $envname
source deactivate
conda list -e requeirment.txt
conda install --yes --file requirements.txt
conda env export > environment.yaml
conda env create -f environment.yaml
conda create -n py36 python=3.6

software

intern net:
firefox http://tntintern/

emial:
thunderbird

linux command

check the size of path:
du -h filepath

screen:
screen -S $Name
screen -ls
screen -r $id
[detached] ctrl+a+d

conda pip env transform

conda环境转移复制和pip包的转移复制

vnc

on servers (cp202-03)
vncserver -geometry 1920x1080 :8
on local (mac)
ssh -fNCT -L 5902:cp202-03:5908 xujing@ssh1.tnt.uni-hannover.de
on vnc view
connect: localhost:5902
kill
vncserver -kill :8

ssh

vscode通过跳板机(堡垒机)连接remote服务器

LocalForward 5902 cp202-03:5908

XuJing's Blog