solution of “CUDA”,“Ninja”
Opened this issue · 0 comments
weikangqi commented
docker is all your need
- first pull docker image
docker pull pytorch/pytorch:1.10.0-cuda11.3-cudnn8-devel
- run docker container
You need to configure it according to your own ???
docker run --gpus all -it --shm-size="32g" --name=SPIN -v /home/???/code:/code pytorch/pytorch:1.10.0-cuda11.3-cudnn8-devel/bin/bash
- in the container
key steps
cd /code/SPIN
pip install -r requirements.txt
chmod 1777 /tmp
apt install ffmpeg libsm6 libxext6 -y
pip install Ninja
python train.py --config ./configs/spin_light_x4.yml