Skip to content

wsl.md

Latest commit

 

History

History
108 lines (85 loc) · 3.85 KB

wsl.md

File metadata and controls

108 lines (85 loc) · 3.85 KB

Note this will install wsl on C disk. make sure your disk have enough space (> 50GB for training a model).
I tested on Windows11 with latest nvidia driver. you don't need to install driver inside wsl.

Step1 install WSL2 on Windows

  1. if you don't have wsl2 on windows, please install wsl2 on windows in command line
wsl --install

You should install wsl2, you can check your wsl version as following:

C:\Users\yourname>wsl -v
WSL version: 2.1.5.0
Kernel version: 5.15.146.1-2
WSLg version: 1.0.60
MSRDC version: 1.2.5105
Direct3D version: 1.611.1-81528511
DXCore version: 10.0.25131.1002-220531-1700.rs-onecore-base2-hyp
Windows version: 10.0.22631.2861
  1. make a workspace dirtory in C:\Users\yourusername as following. For more commands of wsl, please see microsoft's website
mkdir wlsworkspace 
cd wlsworkspace
wsl

after that you should be at this path inside wsl2. (/mnt/c/ in wsl2 is at the path of C: in windows )

yourlinuxname@yourdevice:/mnt/c/Users/yourusername/wslworkspace$

Step2 install cuda toolkit inside WSL2

  1. install cuda toolkit inside wsl2, the instructions from Nvidia are here We select the option for you already, just follow the selected instructions.

  2. add cuda toolkit at PATH by vim editor to the end of file ~/.bashrc as following. You can search how to use vim first.

vim ~/.bashrc

then type in i to insert, move the cursor to the end, paste command export PATH=/usr/local/cuda-11.8/bin${PATH:+:${PATH}} and export LD_LIBRARY_PATH=${LD_LIBRARY_PATH}:/usr/local/cuda-11.8/lib64 tp the end two lines of file, press Esc, then press Shift and : at them same time, type in wq to write the modification (w) and exit (q) the vim.

To make new path effect, please type following in the terminal of wsl:

source ~/.bashrc
  1. check installation of cudatoolkit in wsl terminal by
nvcc --version

following should be returned:

nvcc: NVIDIA (R) Cuda compiler driver
Copyright (c) 2005-2022 NVIDIA Corporation
Built on Wed_Sep_21_10:33:58_PDT_2022
Cuda compilation tools, release 11.8, V11.8.89
Build cuda_11.8.r11.8/compiler.31833905_0

Step3 install miniconda inside WSL2

  1. download minconda in workspace:
wget https://repo.anaconda.com/miniconda/Miniconda3-latest-Linux-x86_64.sh
  1. intall it
bash Miniconda3-latest-Linux-x86_64.sh

add following path to the end of ~/.bashrc.

export PATH="~/miniconda3/bin:$PATH"

To make new path effect and enter conda, please type:

source ~/.bashrc
conda init
conda activate
  1. set conda environment to your work space (suggested, just put conda envs to the workspace incase you want to delete them.)
mkdir /mnt/c/Users/yourusername/wslworkspace/envs
conda config --add envs_dirs /mnt/c/Users/yourusername/wslworkspace/envs

Step 4 clone our repo and follow the commands for linux in readme to setup the environments.

note that building mmcv will take sometime, just leave it there.

bash -i script/setup.sh

Step 5 during training, if you meet error (highly possible) of Could not load library libcudnn_cnn_infer.so.8. Error: libcuda.so: cannot open shared object file: No such file or directory Please make sure libcudnn_cnn_infer.so.8 is in your library path!

follow this comment

suggestions for training on windows without a large gpu memory to avoid CUDA memory error

please set the r to 4(use smaller image size, thanks pablodawson ) in config file or set the duration to smaller values (train with fewer frames).