Weekly updates

• confirmed that the namespace work as expected, including being able to allocate a AMD GPU to a pod
• next is to build a docker image using a trusted gitlab runner and publish it to the wikimedia registry, which is required for running services on kubernetes. we decided to add the research-datasets repo to the list of trusted runners for that purpose and focus on llm inference workloads at first.

Weekly updates

• Progress towards building a rocm wheel for fast attention using gitlab ci, using a custom gitlab runner. Resource challenges around runtime (+10h) and disk requirements (60GB).

Weekly updates:

• the fast attention wheel can be reliably built by CI, and can be for building a docker image for the research-datasets and llmperf repos now.

Moving from the quarterly lane to in-progress as I'm closing the quarterly lane. Please set/update the deadline for the task.