Powering Generative AI Workloads with AMD and Open-Source ROCm
Explore the potential of running generative AI workloads on cost-effective AMD hardware and the open-source ROCm software stack in this 32-minute conference talk. Discover how to implement popular open-source large language and image generation models using ROCm, and learn about the latest features supported, including PyTorch integration, Optimum-AMD, Flash Attention 2, GPTQ, and vLLM. Compare the performance and inference speed of affordable AMD GPUs to their Nvidia counterparts. Gain insights into ROCm's evolution, witness live demonstrations, and receive practical tips for working with AMD GPUs in AI applications. Broaden your understanding of hardware and software options for generative AI implementations beyond traditional CUDA-based solutions.