PaddlePaddle

PaddlePaddle is free and open-source software with sets of the library designed for Machine Learning (ML) to improve its performance. It was exclusively developed by the popular Chinese Search Engine operator Baidu, which is known as one of the widely used search engines around the world. The original developer announced it to be open-source in 2016 under Apache 2.0 License. PaddlePaddle (PArallel Distributed Deep LEarning) got launched, keeping in view the end-user concerns. So it offers a convenient, efficient, flexible, and scalable deep learning platform. Having been developed originally by Baidu scientists and engineers, its primary purpose was to apply deep learning to cope with the functionality flow of multiple products at Baidu, such as NLP (Natural Language Processing), translation Face detection, and image processing (OCR). PaddlePaddle supports vast sets of neural network architectures and several other optimization algorithms. With PaddlePaddle, it has become possible to influence many CPUs/GPUs and machines to agility addition for training, achieving high throughput and performance through an optimized communication medium. Thanks to leveraging RDMA and GPUDirect, a pivotal improvement has been achieved in curtailing training time.

Etymology

The name is technically derived from an acronym for PArallel Distributed Deep Learning (PaddlePaddle). However, not to confuse with English word Paddle, the Baidu administration decided to entitle it with double sound PaddlePaddle.

Framework and algorithms

The framework of the PaddlePaddle construes with a particular focus on training and inference of deep neural networks and graph neural networks. The training process can be parallelized on GPU clusters. PaddlePaddle provides almost 146 algorithms and has advanced more than 200 pretraining models. The framework covers the main categories Computer Vision, Natural Language Processing, Speech Recognition, and Recommender Systems. Furthermore, it also provides experimental features like Quantum Neural Networks (QNNs).

Versions

Since its initial launch, Baidu has periodically developed its versions with the latest additions and updates. Paddle Lite is also an offshoot of PaddlePaddle full pack. Paddle has been serving with its versatility through several versions that were released by its mother corporate Baidu in time to time: V2.3, V2.6, V2.7, V2.8, V2.9, V2.9.1, while 2.10RC is the latest version of PaddlePaddle.

Paddle Lite

Paddle Lite combines tools that can help developers run models on mobile devices, embedded devices, and IoT devices to implement device-side machine learning. PaddleLite is a lightweight version of the framework and intends to support multiple mobile and embedded hardware platforms. It is optimized for ARM CPU, Mali GPU, Adreno GPU, Huawei NPU, and FPGA hardware. Users can also enhance the performance of Ardeno CPUs with Qualcomm-supported modules, thanks to PaddleLite. Now Android systems are more awe-inspiring with Paddle Lite integration.

References

This article "PaddlePaddle" is from Wikipedia. The list of its authors can be seen in its historical and/or the page Edithistory:PaddlePaddle. Articles copied from Draft Namespace on Wikipedia could be seen on the Draft Namespace of Wikipedia and not main one.

📰 Article(s) of the same category(ies)

• SmartCore
• Flokzu
• SOAT (template engine)
• EasyCHAID
• Tiled (software)
• Coherence (UPNP)
• MediaCMS

---