v1.4.0

• Update to support CUDA8 17f6bb1

v1.3.0

• run cuda_init.sh as a systemd oneoff service

v1.2.0

• #3 run cuda_init.sh script before ansible-pull (it only logs an error if nvidia-smi tool is not installed). This is so that the /dev/nvidia files are available before ansible configures and tries to start slurm on the first install
• #4 do not execute one of the handlers (it has a wait_for) if we are running in ansible_pull == 300s speedup of this role.
• Also 897e52f - if there is no /dev/nvidia0 file then we run cuda_init.sh from within ansible.

v1.1.1

• flush handlers: this means the GPU node is rebooted after the cuda role is applied (also all the previous handlers are also applied)

v1.1.0

• new script: cuda_init.sh which is added to rc.local - sets persistence and compute mode
• changed sudo to become (ansible2)

v1.0.1

quote variabels used in loops - ansible2

v1.0.0

Fix reboot on installation.

v0.9.0

wrong syntax