NVIDIA Tensor RT 3, the latest reasoning software, will be deployed on a Deep learning platform based on Tesla V100 GPU acceleration
On October 31, Meituan cloud ARTIFICIAL intelligence Summit was held in Beijing. At the conference, meituan cloud officially announced that it has reached a cooperation with NVIDIA, and will launch GPU cloud host based on NVIDIA Tesla V100 GPU accelerator in China, and provide high-performance computing services for meituan cloud AI enterprise users such as face++.
Based on the excellent performance of Tesla V100 GPUs in performance, power consumption, cost and other aspects, Meituanyun will also launch the deep learning platform based on Tesla V100 in China for the first time, providing model training, reasoning and other functions for the majority of developers to increase the usability and ease of use of the deep learning platform.
NVIDIA released the latest version of TensorRT 3 inference software, will also be deployed on Meituan Cloud Tesla V100 GPU cloud host image, greatly improve the inference transmission capacity of deep learning framework, for image and language recognition, natural language processing, visual search and personalized recommendation and other fields of AI enterprises. Provide more efficient AI services.
It is reported that the meituan cloud deep learning platform is free for developers to use.
Meituan has previously launched GPU cloud hosts based on NVIDIA Tesla M60 and P40, and provides services for AI research projects such as Beijing Language and Culture University and The Chinese Academy of Sciences, as well as AI companies such as Rokid.
Providing high-performance GPU cloud services based on Tesla V100 to accelerate computation-assisted deep learning
At this year’s GTC conference, NVIDIA announced the Tesla V100, the most powerful GPU ever. The Tesla V100 boasts more than 21 billion transistors, 15TFLOPS of single-precision floating-point computing power, three times the performance of deep learning training, and 100 times the throughput of reasoning as a CPU server. The high-performance Tesla V100 can help academic researchers and ai application r&d companies break through the original computing bottleneck, accelerate their own research progress and promote technology commercialization.
According to IT Orange, there are 1,201 AI start-ups in China, providing services in medical, education, finance, retail, logistics and other fields. As it takes a long time for artificial intelligence enterprises to realize business value from technology research and development, product transformation, enterprises mainly consider three factors when choosing GPU: floating point computing capability, power consumption and cost.
Typically, the floating point computing power of a GPU is proportional to its power consumption, while the innovative energy saving mode of the Tesla V100 GPU can increase the computing power per rack by up to 40% without increasing the budget cost of the data center. In addition, during operation, the Tesla V100 can generate only 50% power consumption while achieving 80% of the performance operating state.
NVIDIA GPU expands artificial intelligence track with AI explosion
High-performance computing power is a necessary condition for the development of the ARTIFICIAL intelligence industry. With the elevation of artificial intelligence to a national level, the whole artificial intelligence market has ushered in a development opportunity. As the preferred computing platform for ARTIFICIAL intelligence, NVIDIA GPU is increasingly in demand by high-performance computing, artificial intelligence and deep learning users.
At present, domestic Internet companies, cloud service manufacturers and server manufacturers are increasing the number of GPU-based platforms for different application scenarios, transforming data centers, or deepening cooperation with AI chip manufacturers to cope with the growing demand for HIGH-PERFORMANCE computing and artificial intelligence applications.
Meituancloud first launched the Tesla M60 GPU cloud host in March this year, and then launched the Tesla P40 GPU cloud host and other computing services suitable for different scenarios. With the upcoming Tesla V100 GPU cloud host, the computing capacity of Meituancloud GPU cloud host has been improved again. According to Meituuan cloud, the company will build the first V100-based GPU cluster to accelerate deep learning in the future.
Meituan Cloud is the public cloud platform of Meituan Dianping, The largest Internet + life service platform in China. It started by serving Meituan Dianping and helped Meituan Dianping realize 100% cloud of its core business. Meituan Dianping is also the first Internet company in China to achieve 100% cloud. Meituan cloud currently has 40,000 enterprise and individual users, including sogou, yitu, Rokid, face++, jumei youpin, renyendai, huazhu hotel, Chinese Academy of Sciences, Beijing language and culture university and other partners and customers.