AI BriefingHugging Face-AMD Partner to Accelerate AI Models
Key point
Hugging Face and AMD have formed a partnership to optimize transformer model performance on AMD CPUs and GPUs.
Details
Hugging Face is bringing on AMD as an official partner of the Hardware Partner Program, beginning a collaboration to maximize the performance of state-of-the-art transformer models on AMD hardware platforms.
Key Supported Hardware:
- GPU: Enterprise Instinct MI2xx/MI3xx and consumer Radeon Navi3x series
- CPU: Client Ryzen and server EPYC processors
- Accelerator: Alveo V70 AI accelerator
Supported Models and Frameworks: It supports major generative AI models such as BERT, GPT-2, LLaMA, BLOOM, and StarCoder, as well as computer vision models like ViT and CLIP. The goal is to provide optimized performance across PyTorch, TensorFlow, and ONNX Runtime environments.
Future Plans:
The plan is to seamlessly integrate the AMD ROCm SDK into the transformers library, and to build an AMD-specific Optimum library that lets users leverage AMD hardware with minimal code changes.
This summary was generated automatically by AI. Check the original for the author's claims and context. Copyright belongs to the original author.
Our guide explains how the AI works. Report summary errors, attribution issues, or removal requests via Contact.