Huawei MindSpore computing platform is similar to Google TensorFlow. But the latter has the advantage of being an open source platform. Following in its competitor's footsteps, Huawei has also made Mindspore open source. The company announced this during the Huawei Developer Conference Cloud 2020 event.
Chinese tech giant Huawei for the first time MindSpore platform for AI computing in August 2019 along with its custom Ascend 910 processor. MindSpore has three main goals: ease of development, efficient code execution, and the ability to adapt to any scenario, respectively.
Since privacy has become an issue in the modern world, a lot of attention in the development of MindSpore was given to data protection and the lack of direct access to personal information. The MindSpore infrastructure can be used in all scenarios, on all devices: both on endpoints like smartphones and in the cloud.
Because MindSpore requires 20% fewer lines of core code than other platforms for typical NLP (Natural Language Processing) neural networks, the company claims development efficiency increases by at least 50%. The Huawei MindSpore infrastructure supports not only its own nanoprocessors like the aforementioned Ascend 910, but also other processors and graphics accelerators available on the market.
Source: 3dnews.ru