Intel and Baidu Team Up to Optimize Custom AI Training Chip
Intel and Baidu announce a partnership to optimize Intel's upcoming Nervana Neural Network Processor for Baidu's PaddlePaddle deep learning framework. The collaboration focuses on hardware and software enhancements to improve large-scale distributed neural network training.
Intel and Baidu announce a new partnership at Baidu's Create conference in Beijing to collaborate on Intel's upcoming Nervana Neural Network Processor for training, also known as NNP-T. This custom chip is built specifically to handle the demanding task of training neural networks for deep learning at scale.
The joint effort focuses on optimizing both the hardware and software sides of the NNP-T to work seamlessly with Baidu's PaddlePaddle deep learning framework. This collaboration builds on Intel's existing work to ensure PaddlePaddle performs well on Intel Xeon Scalable processors, with a specific emphasis on improving distributed training capabilities.
Developed by the Intel AI group led by former Nervana CEO Naveen Rao, the NNP-T is designed solely for training AI models, unlike its counterpart the NNP-I which handles inference. The new "Spring Crest" generation of the NNP-T targets production availability this year, following the first-generation chip's use as a software development prototype.