Chinese search engine Baidu formally announced plans last month to open source its deep learning platform named PaddlePaddle, and now the project is going full-steam at GitHub.
Since its original announcement on September 1, Baidu has become the latest tech company which to make its artificial intelligence technologies open source. Google, Facebook, and IBM have previously made their artificial intelligence and deep learning technologies open source. Since that announcement, its project hosted on GitHub has had many updates.
The predecessor of PaddlePaddle is Baidu's self-developed deep learning platform Paddle, which is also known as Parallel Distributed Deep Learning. This platform has been used internally used by Baidu's engineers for research and development purposes.
PaddlePaddle covers technologies of various sectors and businesses, including search, image identification, speech semantic recognition, emotional analysis, machine translation, and user portrait recommendation. On September 1, 2016, Baidu's chief scientist Andrew Ng plans announced to open Baidu's deep learning platform to the public for the first time and named it PaddlePaddle.
The platform supports widely-used neural network architectures, such as convolutional neural networks and recurrent neural networks, and can scale to a large number of GPUs or CPUs on multiple machines. To take advantage of the power of heterogeneous computing, optimization occurs at different levels of the platform including computing, memory, architecture and communication.
At present, PaddlePaddle has been used by over 30 major products and services of Baidu.