Shulou(Shulou.com)11/24 Report--

November 2, 2023 is the last day of the three-day Cloud Congress. But for developer Tan Yaohua, it was the first day that he and his team won the first prize in the Intelligent Operation and maintenance Competition and the second prize in the Intelligent creation Competition.

One key word was repeatedly mentioned at this year's Yunqi conference: developer. Zhou Jingren, CTO of Aliyun, believes that the big model is still in the initial stage of evolution. Whether you want to develop it yourself and what kind of model to use should be determined by the effect. You should stand from the developer's point of view, give developers tools and choices, and let them choose their own models.

The first data Insight Innovation Challenge, sponsored by Ali Yun and jointly organized by Ali Yun Tian Chi platform and Ali Yun storage team, is a technical competition for developers. After nearly half a year of fierce competition, a total of 10 teams broke out of the siege from more than 5000 teams and came to the final scene of Yunqi.

(award photo)

Code achievement Wanshiji sand Zhenhai, the dream is always high-spirited Lingyun. For developers, the data Insight Innovation Challenge not only provides a platform to show their technology and wisdom, but also provides new ideas and directions for innovation and development in the field of AIGC and AIOps.

First, "AI + creation"-- telling stories with images to make photos warm.

AIGC represents the transition of AI technology from perceiving and understanding the world to generating and creating the world, and is promoting artificial intelligence to usher in the era of intelligent creation. In recent years, the parallel development of Wen Sheng Wen, Wen Sheng Picture and Wen Sheng Video / Picture Video is obviously inseparable from the big model behind it. It is worth mentioning that the large model has significantly lowered the threshold for developers. Xiaoyi, head of Aliyun's basic products division, firmly believes that with the support of the vast number of developers, Aliyun can build a more perfect and complete technology product ecology.

The contestants of the intelligent creation competition rely on the technical capabilities of the online disk and photo album service PDS to "summarize" the pictures in the album, and use the current mainstream AIGC technology to do picture style transfer, portrait style transfer, image editing, etc., resulting in oil painting style, pencil style and other styles of "recall story" album and video, showing the beauty of the combination of data and art.

Fan Xianghang, a contestant in Maogu Love Photography, the champion team of the intelligent creation competition, is a professional risk control algorithm engineer and a photography enthusiast. He laughs that he is a photo enthusiast who has participated in the Tianchi algorithm contest more than 20 times. In the process of the rematch, he not only gave full play to a photographer's aesthetic understanding of photos, but also made use of graphic text model, picture quality evaluation model, text vectorization model, hierarchical clustering algorithm and other means. finally, the style transfer of pictures, title generation, music type selection and automatic video generation are realized.

During the competition, Xiong Cheng, a new data team from Shanghai University of Science and Technology, used StableDiffusion in AIGC technology combined with ControlNet to customize image style editing according to user needs; and used BLIP model to identify multi-image semantics and judge image emotion to determine the needed soundtrack; at the same time, he also used image text combined with GPT2 to expand and write stories, which were added to the album video.

Li Wenzhao, a representative of the final judges and a senior technical expert from Aliyun, commented that through the presentation of the entries, the organizers saw the developers' ability and innovative thinking in the process of problem understanding, solution design and implementation. Contestants can not only skillfully use data analysis and machine learning algorithms, but also combine their own experience and expertise to make use of Ali cloud storage products and services, and put forward excellent solutions.

Second, "AI + operation and maintenance"-- automatically locate faults and improve efficiency for "engineers"

As the business goes to the cloud, the operation and maintenance is also gradually "cloud", and the architecture based on micro-services plays a more and more important role in business development. In micro-service system, the overall delay and success of business request (RPC) are closely related to the dependent micro-service invocation. When the business request slows down and fails online, engineers often need to find the key node and reason in each calling node and eliminate it.

The team of the Joint Laboratory of Cloud observation of Zhejiang University believes that in the field of information technology, cloud native and micro-service architecture has become the main trend, providing flexibility and scalability, but as the micro-service system expands, maintenance and troubleshooting become complex. Business requests in multiple micro-services and distributed environments increase the difficulty of troubleshooting. Based on this actual situation, the team proposed an automatic fault location algorithm based on statistics, which aims to automatically locate the root cause of slow requests through data analysis and algorithms. The method includes four steps: sampling, preprocessing, statistics and root cause analysis. the team members discuss different anomaly detection algorithms in detail, among which the boxplot algorithm shows higher performance.

The team from Sun Yat-sen University, a team, has been deeply cultivated in the field of intelligent operation and maintenance for many years, and the algorithm results have landed in the cooperative enterprises. In order to diagnose high latency Trace data and wrong Trace and get the corresponding root cause, a traceability-oriented

The processes of monitoring data analysis, anomaly detection based on hierarchical modeling, root cause location, and metamorphosis test feedback based on collaborative analysis of indicators and tracking data are analyzed and dealt with the root cause positioning challenges of abduction, modeling and feedback.

Jian Zhi, senior technical expert and head of Aliyun log service technology, said that it is very important to build a systematic observation system through the organic fusion of multi-dimensional observation data link Trace, time series indicator Metric and log details Log, while providing a wealth of observation means from business application perspective (including dependency analysis, performance analysis, fault troubleshooting and root cause location). It is very important to achieve the transition from passive perception to active observation.

Conclusion: data-driven innovation in view of the future. Doing a good job of data insight is a necessary condition for learning from the past. Through this competition, we have seen the talent and creativity of the contestants, as well as the application potential of data insight in various fields. In the future, Aliyun will join hands with developers to further improve relevant technologies and tools, provide better products and solutions, and make the application of data insight more extensive and in-depth.

Attached: list of winners of data Insight Innovation Challenge

Intelligent operation and maintenance race:

First prize: Pikachu pickup

Second prize: a team, the Joint Laboratory of Cloud observation of Zhejiang University

Third prize: hoarz, reasonable imagination team, Changjie Mathematical Intelligence team

Intelligent creation Competition:

First prize: Mao Gu loves to take pictures

Second prize: Pikachu pickup, brave Niuniu is not afraid of difficulties

Third prize: data freshmen, competition team, Changjie Mathematical Intelligence team

Welcome to subscribe "Shulou Technology Information " to get latest news, interesting things and hot topics in the IT industry, and controls the hottest and latest Internet news, technology news and IT industry trends.