Hi, Jack!
Are you referring to Merlin as it relates to nVidia, eg:
https://github.com/NVIDIA-Merlin/Merlin? If so, the answer may seem a little disappointing, although I'm not specifically familiar with Merlin and hadn't seen any prior requests. However, with the newer Power10 launch, the Power brand chose to move away from a training focus for AI and ML which was focused on GPUs and instead to a more enterprise focused AI model which makes use of the Power10 intrinsics around mma and related accelerations to do more AI and ML processing rather than training/learning. As I understand it, the rate of scaling GPUs to the rate of scale of Power10 made GPUs more cumbersome and inefficient to make use of, hence the focus on enterprise AI and the elimination of work around embedded GPUs and nVidia solutions in general.
Are there other workloads that Merlin is suited for that do not relate to GPUs or if there is any way my answer is incomplete let me know, but thanks for the (hard! ;-) question!
------------------------------
Gerrit
------------------------------
Original Message:
Sent: Wed September 21, 2022 12:25 PM
From: Jack Woehr
Subject: OpenShift and Merlin on Power
Who's doing Merlin on Power?
How's your OpenShift experience in that environment?
------------------------------
Jack Woehr
IBM Champion 2021 -2022
IBM Qiskit Advocate
------------------------------