Skip to content

Latest commit

 

History

History

docker

Folders and files

NameName
Last commit message
Last commit date

parent directory

..
Dockerfile
Dockerfile
 
 
README.md
README.md
 
 

README.md

Run the following command to build the docker for tensorcircuit at parent path:

sudo docker build . -f docker/Dockerfile -t tc-dev-jupyter

Run the docker container by the following command:

sudo docker run -it --network host -e LD_LIBRARY_PATH=$LD_LIBRARY_PATH:/usr/local/cuda-11.0/targets/x86_64-linux/lib -e PYTHONPATH=/app -v "$(pwd)":/app  --gpus all tc-dev-jupyter

To utilize tensorflow + cuda, we need to further run the following command within runtime docker:

export LD_LIBRARY_PATH=$LD_LIBRARY_PATH:/usr/local/cuda-11.0/targets/x86_64-linux/lib

export TF_CPP_MIN_LOG_LEVEL=3 maybe necessary since jax suprisingly frequently complain about ptxas version problem. And export CUDA_VISIBLE_DEVICES=-1 if you want to test on CPU.