[Large model technology at the terminal] OSC Source Foundation · Shenzhen Station · Phase 103

Status: Activity ended

Time: 2024-05-26 14:00 to 18:00

Location: Entrance 15, Gangxia North Subway Station, Futian District, Shenzhen, Guangdong

cost: Free Admission

Activity registration

Real name:

phone number:

corporate name:

Company Position:

Activity review https://mp.weixin.qq.com/s/xNNUnphO2R_SiRG8n96Sbw

🌟 theme introduce

This technology salon will deeply discuss the application and development of the big model technology in the field of intelligent terminals, as well as some key technologies that support the development of the big model technology, such as the optimization of programming languages and developer tools for the generation of end code, the improvement of serverless reasoning engine for the reasoning efficiency of the big model, and so on.

In order for OSCer to have a better participation experience, we have carefully selected a booming hall, which integrates learning, making friends, entertainment and leisure! Call your developer friends to play!

Snacks+drinks, unlimited self-help throughout the whole process, as well as board game props such as werewolf killing and the judge's play

🌟 Introduction to speakers and topics

Speaker: Yao Yifan, Senior Engineer of MindSpot Large Model

Topic: Sharing of training technology of Shengsi MindSpot large model

Topic introduction:

Large scale AI driven by large models, big data, and large computing power is becoming a breakthrough for the next generation of AI, and is also one of the most promising technology directions for general strong AI. Shengsi MindSpot has built key technologies such as multi-dimensional hybrid parallelism, multi-dimensional memory optimization, and graphic and computational integration, which supports the training of 20+major models. In 2021, we cooperated with universities and scientific research institutes to make collaborative innovation, and released Pengcheng Pangu NLP model, Zidong Taichu multi-modal model of Chinese Academy of Sciences, Pengcheng Shennong protein structure model, and Wuda Luojia remote sensing model. This time, we share the latest progress and achievements of the large model based on MindSpot, introduce the key technologies of the whole process of large-scale AI training, fine-tuning, compression and deployment of MindSpot, how to use these key technologies to train the large model on the Shengteng large cluster, and how the large model on MindSpot can help industrial development.

Speaker: Zheng Chongyang - Senior Product Manager of iFLYTEK Open Platform

Topic: Spark model+application innovation of intelligent hardware

Topic introduction: The smart hardware industry is facing the pain point of product homogenization and lack of innovation, resulting in fierce internal volume and thin profits. The emergence of AI model will greatly promote the intelligent process of thousands of industries, and will also greatly enrich the application innovation of intelligent hardware. This topic will share applications in watch, keyboard and mouse and computer PC industries, hoping to jointly explore deeper scene innovation and more industrial applications.

Speaker: Fei Haoxiang, Senior Engineer of IDEA Basic Software Center

Topic: End code generation: take MoonBit as an example

Topic introduction:

As the main tool of developers, IDE can respond to users' input in real time, work offline, and intelligently prompt and complete the retrieval code base. This sharing will take MoonBit as an example to introduce the advantages of the end model in code generation and the optimization of the MoonBit IDE for the end code generation.

Speaker: Lin Jiazhen - Gitee AI expert consultant, High Performance Computing Institute of Tsinghua University

Topic: Use Serverless inference engine to improve the reasoning efficiency of large models

Topic introduction: The native application of the big model is exploding, and the demand for reasoning computing power may reach tens of times of the demand for training computing power in the future. According to Gartner data, 70% of AI applications will be developed based on Serverless technology in 2023. In actual production, more and more AI businesses, such as autopilot and NLP, are also turning to containerized deployment. Is there a chance to help the big model reduce costs and increase efficiency? What challenges have you encountered in this process?

🌟 "Welcome" activity is open!

📒 Tips for winning prizes:
1. Participate in Shenzhen Yuanchuang Fair
2. Invite developers and friends to attend the conference
3. Open source China exquisite peripheral get~[wow] [wow]

✍️ Specific rules:
1. Send the activity registration link to the developer friends and invite TA to participate in the conference
2. The invitee shall fill in the name of the inviter in the "Name of the inviter" column of the registration form
3. On May 26, when more than three invitees signed in at the event site, the inviter could get a T-shirt or an open-source Rubik's Cube around OSC

🌟 Activity communication group

🌟 Prize Pool

🌟 cooperative partner

🌟 Volunteer Recruitment

If you want to enter the OSC editorial department, if you want to know how a technical salon is organized, if you want to get the beautiful surroundings of OSC, come and join us and become a volunteer on the spot!!! (Please add WeChat: 18655807197)

🌟 About Source Foundation

OSC Source Foundation is hosted by the Open Source China Community (oschina.net), which focuses on open source and innovative technology salons. Adhering to the purpose of "freedom, openness, and sharing", Source Innovation Association gathers the best technical resources and industry cases, talks with the best technology leaders, and brings the latest open source technology, cutting-edge technology perspective, and landing practice experience to the majority of developers.

🌟 About Open Source China

OSCHINA Founded in August 2008, it has now established a fairly complete open source software classification database, including nearly 100000 world-famous open source projects, involving hundreds of different classifications. Around these open source projects, OSCHINA provides a technical platform for Chinese developers to share and exchange the latest open source information, software update information, technology. In 2013, OSCHINA established the code hosting and DevOps platform "Code Cloud Gitee" to provide developers with team collaboration, source code hosting, code quality analysis, code review, testing, CI/CD, code demonstration and other functions.

After more than ten years of deep cultivation in the open source field, and the combination with China's local open source environment, it has promoted the rapid development of China's open source field. OSCHINA has now developed into a well-known open source technology community in China, with 6 million active developers. It has long been committed to promoting the application and development of domestic open source software, improving the local open source capability, and providing support for the optimization of the open source ecological environment.

comment

Ctrl/CMD+Enter Comment

People
Nature
Objects
Places
Symbols