What if the computer cannot run the assignment that requires great computing capacity? How to find solutions for slow transmission while AI model needs a large amount of data? What if the experiment cannot be carried out while the software downloaded are incompatible with each other? ...... There have been many new troubles arising in teaching and research in the AI era. To deal with these, Tianjin University builds an AI cloud platform to intelligently transform the classroom and equip a digital brain for education and research.
Tianjin University’s College of Intelligence and Computing was established in accordance with the national policy to enhance the construction of a new generation of Artificial Intelligence, while advantageous resources in Tianjin University in high-tech IT fields such as Cognitive Computing, Data Science, Smart Technology and Applications have also been brought together. Recently, the college introduced Artificial Intelligence into teaching, and provided a unified computing platform named as Tianjin University Cloud Platform for Artificial Intelligence Research, for the School of Computer Science and Technology, the School of Software Engineering, the School of Artificial Intelligence and the School of Cyber Security.
The AI platform integrates high-performance computing clusters, GPU clusters, virtualization clusters and a cloud platform together. Under the interconnection of high-speed 10 Goptical fiber networks, high-performance computing clusters can support computing in batch processing mode, and GPU clusters can support computing in AI reasoning and calculus, while virtual machines on cloud platforms can be used for program debugging and small-scale experimental teaching. The AI platform fully covers computing needs required by future possible projects.
Students gave favorable feedbacks (5 star) after using the AI cloud platform.
"In my lip reading research, there were more than 500G of data. In the past, due to internal storage limits, much data has to be deleted and then the rest then can be uploaded in batches. With this AI platform, I can upload large files without bandwidth or storage restrictions, and 500G data can be uploaded successfully in two hours."
"Although there are many high-performance GPUs in the lab, students and tasks requiring those resources are no less, leading to vey slow model training process. It used to take about one week to finish an experiment. It is much faster now to use AI platform for model training, and the similar experiment can be completed in 1-2 days. This is very helpful for model fine-tuning."
"In the past, if you want to check the intermediate results of the model, you need to copy the pictures from the server to the local PC first. Now you can just check the results in real time directly on the AI platform."
Wang Jianrong, Associate Professor from Tianjin University’s College of Intelligence and Computing further elaborated the platform. "For senior students and graduate students who worked in labs every day, their project progress is closely related to the experimental environment. Open source software downloaded by themselves often faces problems such as system incompatibility, etc. This AI platform provides basic software environment for deep learning training and reasoning, data mining, scientific computing, as well as task scheduling systems that can support GPU. This will provide students with an experimental environment that is in line with the industry. In addition, all types of resources can also satisfy the needs of students from different departments. "
With the support of this AI platform, Tianjin University’s College of Intelligence and Computing can not only successfully launch major national science and technology projects and various projects on the provincial and ministerial levels, but also further promote Industry-university-research transformation.
It is learned that this platform not only fully supports teaching and research related to artificial intelligence, big data processing and analysis, high-performance computing, etc. in the College of Intelligence and Computing, but also expands cloud infrastructure resources through professional operations and provides support for other schools at Tianjin University as well.
By the College of Intelligence and Computing
