With an aim to reach more developers and advance computer applications, Intel has launched a new open source deep learning library called BigDL. The new development is specifically designed to let developers write deep learning applications for Apache Spark cluster computing framework.
Intel’s BigDL runs in the Databricks Spark Platform to help building of advanced applications as standard Spark programs. A big advantage of the latest library is its operative behaviour as a universal data analytics platform using Spark or Hadoop for data storage, mining and processing. There is also support for engineering, deep learning and machine workloads.
BigDL is based on Intel MKL and multi-threaded programming in Spark task. Furthermore, the library supports deep learning and scientific computing using Tensor and some high-level neural networks.
Developers can load their pre-trained Torch or Caffe models in Spark programs to work on some major projects by simply using BigDL. Moreover, the new release is seemed to be the part of Intel’s overall strategy for artificial intelligence (AI).
You can access the BigDL code directly from its GitHub repository. It also includes necessary documentation such as tutorials, examples and programming guide to help you develop your next deep learning offerings.