Install Geti 2.13.1 on an environment with clinfo not installed.
The result: cpu is set as device for running training, which ends up with out of memory errors when running training.
Workaround:
- either install clinfo prioir to running the Geti installer, or
- when Geti is installed, edit the config mam and set the correct values as shown below:
sudo kubectl edit cm accelerator-configuration -n impt
and then, set:
accelerator_name: gpu.intel.com/xe
accelerator_type: gpu
or
accelerator_name: gpu.intel.com/i915
accelerator_type: gpu
depending on the driver used for your Intel card.