Building on the architecture of Gaudi, which launched first on the AWS EC2 cloud in the DL1 instance, and on-premises via the Supermicro X12 Gaudi Training Server, Gaudi2 brings a new level of performance and efficiency to deep learning in the data center and cloud. Gaudi2 is supported by the SynapseAI software suite, which is integrated with TensorFlow and PyTorch frameworks. We offer reference models, tools, and how-to guides on Habana’s GitHub and Developer Site. The main benefits that current customers of our first-generation Gaudi see are the price-performance advantages relative to GPU solutions for popular vision and language models. These enable customers to train more and pay less, and this way, accelerate time-to-market with their model training.
