churchlab / ml-ami-builder

Packer scripts to build nvidia-enabled AMIs
19 stars 4 forks source link

Clean up unused install files #5

Closed glebkuznetsov closed 7 years ago

glebkuznetsov commented 7 years ago

Eventually trying to cleanup the nvidia install where we are having an issue with the cuda driver that randomly starts happening if docker is not safely stopped. This commit is just doing a bit of cleanup before we get to those real fixes in later commits.

Tested by:

1) Used packer to build ami:

    (local)$ packer build \
        -var 'aws_access_key=<ACCESS_KEY>' \
        -var 'aws_secret_key=<SECRET_KEY>' \
        gpu-packer.json

2) Created GPU instance (p2.xlarge) from AMI.

3) Cloned https://github.com/churchlab/mlpe-gfp-pilot.

4) Built GPU docker image:

    (aws)$ docker build -f docker/Dockerfile.gpu -t mlpe-gfp-pilot-gpu .

5) Ran a docker container from the image:

    (aws)$ nvidia-docker run -it -p 8888:8888 -p 6006:6006 -v ~/notebooks:/notebooks mlpe-gfp-pilot-gpu

Confirmed no error.

grinner commented 7 years ago

This seems cool to do since we are all in on docker now