Intel and Hugging Face Partner to Democratize Machine Learning Hardware Acceleration
Hugging Face Blog
Intel joined Hugging Face's Hardware Partner Program to develop optimization tools that reduce latency and improve performance for Transformer models running on Intel platforms. The collaboration achieved single-digit millisecond latency for DistilBERT on Intel Xeon Ice Lake CPUs and Habana Gaudi accelerators deliver up to 40% better price-performance than GPUs. The Optimum Intel open-source library now allows machine learning practitioners to apply quantization, pruning, and other optimization techniques to Transformer models with minimal code changes.