Video: Pretraining on AMD MI300X using ScalarLM (feat. Greg Diamos, MLCommons)

How ScalerLM is simplifying large-scale LLM workflows on next-gen AMD hardware

What if your next model training loop could run 100B-parameter LLMs without wrestling separate stacks for training and inference, and without vendor lock-in? In this Stack Session, Greg Diamos, Founder of MLCommons, introduces ScalerLM: an open-source platform built for unified LLM training and inference across AMD and NVIDIA GPUs.

Drawing from his experience leading CUDA software at NVIDIA and pioneering deep learning infrastructure at Google Brain and Baidu, Greg walks through why scale and openness matter more than ever, and how ScalerLM aims to become the go-to stack for reasoning-capable AI.

This talk was recorded during Stack Session 02, part of TMLS’s free professional webinar series for AI practitioners.

Presentation Highlights

This talk is for engineers, researchers, and infra teams building or scaling LLM workloads on multi-GPU setups.

  • Why unifying training and inference stacks enables faster iteration and richer LLM applications
  • How ScalerLM fits into the open ecosystem alongside HuggingFace, Megatron, and VLLM
  • What early performance benchmarking reveals about AMD’s MI300X for deep learning workloads
  • A walkthrough of reasoning loops using ScalerLM with just 19 lines of code
  • Open benchmarks and reproducible baselines for MI300X memory bandwidth and inference throughput

About The Speaker

Greg Diamos is the Founder of MLCommons and a core contributor to ScalerLM. He previously helped lead CUDA architecture at NVIDIA and scaled early deep learning clusters at Google Brain and Baidu.

His work focuses on high-performance AI systems, benchmarking, and open infrastructure.

3 Days of Context, Insights, & Connections

The 6th annual MLOps World | GenAI Summit is taking place October 7–9, 2025 at the Austin Renaissance Hotel.

Don’t miss this chance to accelerate and de-risk your AI/ML, agentic, and infrastructure outcomes through cutting-edge strategies, real-world case studies, technical deep dives, and hands-on workshops. Every presentation is hand-picked by a committee of fellow AI practitioners to spotlight where the line of excellence is, right now.

The experience includes a vibrant expo, where attendees shift from focused learning to active participation by engaging in Brain Dates, Community Stage, Startup Zone, interactive demos with top vendors.

MLOps World | GenAI Summit is a high-impact way to learn, connect, and elevate your team, projects, and career.

Early Bird tickets are on sale now and offer 15% savings when you register in advance.

Table of Contents
Share This Post