Adding this one on here because it was so quick to bang out.
Note to users: You will need to specify entrypoint_num_gpus upon job submission w/ the Ray client if you want to run a GPU-powered lm-harness evaluation now. It can no longer be specified on the config YAML file.
Adding this one on here because it was so quick to bang out.
Note to users: You will need to specify
entrypoint_num_gpus
upon job submission w/ the Ray client if you want to run a GPU-powered lm-harness evaluation now. It can no longer be specified on the config YAML file.