PhariaFinetuning service resource requirements

This article describes the requirements for your finetuning service resources.


Pod scheduling

For each job that is submitted to the PhariaFinetuning service, the following additional pods are scheduled:

2 pods with:

  • 1 Nvidia GPU with at least 80GB of GPU RAM (e.g. NVIDIA A100, H100/200, GH200, B100/200)

  • 120 GB of RAM

  • 14 CPUs

  • 200GB of disk space

1 pod with:

  • 2 CPUs

  • 14GB of RAM

  • 50 GB of disk space

Resource availability

In the event that the above resources are not available, finetuning jobs will be pending.

If more than one job is submitted, jobs run in parallel if enough resources are available. Otherwise, jobs are queued.