Nvidia launches software to speed up deep learning networks

The company released software to spread work around to multiple GPUs to accelerate the training of deep learning networks.

Nvidia on Tuesday launched software that uses its graphics processors to speed up the time it takes to train deep learning networks.

Deep learning networks serve as the base for artificial intelligence and its ability to process data and learn.

Specifically, Nvidia released new versions of its DIGITS Deep Learning GPU Training System and its CUDA Deep Neural Network library.

In a blog post, Nvidia said that the software will accelerate deep learning projects and product development. Since deep learning networks will process information faster, researchers and data scientists can create more sophisticated models.

The second version of DIGITS moves neural network training across multiple GPUs. The CUDA library optimizes data storage in GPU memory and be optimized in popular deep learning frameworks.


Yahoo's Flickr and Baidu Research have been among the early customers trying out the software.

The preview of DIGITS is available today.