## Overview Offer GPU-enabled virtual machines on GitLab SaaS Shared Runners to support the [ModelOps strategy](https://about.gitlab.com/direction/modelops/) of empowering GitLab customers to build and integrate data science workloads within GitLab. ### Naming convention As per [this discussion](https://gitlab.com/groups/gitlab-com/gl-infra/-/epics/805#note_1313656654) our GPU-enabled runners should follow the following naming convention `saas-linux-{size}-{platform}-gpu-{gpu_type}` **size**: `small` = 2vCPU, `medium` = 4vCPU, `large` = 8vCPU, `xlarge` = 16vCPU **platform**: `amd64`, `m1`, `arm` **gpu-type**: `lite` = T4, `standard` = P4, `premium` = A100, V100, `ulta` ... ### Planned machines | Runner Tag | vCPUs | Memory | Storage | GPU | GPU Memory | |----------------------------------------|-------|--------|---------|--------------------------------|------------| | `saas-linux-medium-amd64-gpu-standard` | 4 | 16 GB | 50 GB | 1 Nvidia Tesla T4 (or similar) | 16 GB | | `saas-linux-large-amd64-gpu-standard` | 8 | 32 GB | 100 GB | 2 Nvidia Tesla T4 (or similar) | 32 GB | | `saas-linux-xlarge-amd64-gpu-standard` | 16 | 64 GB | 200 GB | 4 Nvidia Tesla T4 (or similar) | 64 GB | | `saas-linux-medium-amd64-gpu-premium` | 4 | 16 GB | 50 GB | 1 Nvidia Tesla V100 (or similar) | 16 GB | | `saas-linux-large-amd64-gpu-premium` | 8 | 32 GB | 100 GB | 2 Nvidia Tesla V100 (or similar) | 32 GB | | `saas-linux-xlarge-amd64-gpu-premium` | 16 | 64 GB | 200 GB | 4 Nvidia Tesla V100 (or similar) | 64 GB | ## Business outcomes {placeholder for business outcomes} ## Success metrics {placeholder for success metrics}