v2.3.0

• #64 : Allow specifying package variant for drivers from the Ubuntu repositories
• #65 : Add Ubuntu 22.04 to molecule tests, verifying support for Ubuntu 22.04

Thanks @guenhter for contributing these PRs!

v2.2.1

Update to new NVIDIA signing key #59

v2.2.0

• #57 : Fixes for failing Molecule tests
• #58 : Update default driver branch to 510

v2.1.0

• #46 : Update metadata to include support for EL8
• #48 : Make optional the addition of yum/apt repos
• #51 : Explicitly blacklist nouveau when installing from CUDA repos
• #52 : Explicitly add support for CentOS/RHEL8 to README
• #53 : Align default driver branch across all installs and update to r470
• #55 : Add Molecule CI test for nvidia driver role

v2.0.2

Fixes issue with CentOS 8 EPEL repo GPG key (#45)

v2.0.1

Incorporates various bug fixes and improvements

PRs #30, #31, #32, #33, #38

v2.0.0

On Ubuntu: Use NVIDIA server drivers from Canonical by default (#24) and provide an option to use drivers from the NVIDIA CUDA repository.

Note: this is a change from the previous behavior, where the NVIDIA CUDA repository was used by default. This behavior was changed to track the -server driver branch by default, as this branch tracks a more stable branch for datacenter use.

To use the NVIDIA CUDA repository instead of the Canonical -server packages: set nvidia_driver_ubuntu_install_from_cuda_repo: yes.

For more detail on configuring the package behavior: see the Ubuntu driver configuration variables.

The behavior for Red Hat-based distributions is unchanged.

v1.2.2

Bugfix for driver installs on Enterprise Linux distributions (#21)

v1.2.1

Bugfix from @ClementGautier for installing specific version of NVIDIA driver on RHEL. (#13)

v1.2.0

Merge HTTP proxy support from @phogan-nvidia