We are thrilled to announce that our team will be presenting a talk and a tutorial on the topic "Colossal-AI: Scaling AI Models in the Big Model Era" at the virtual NVIDIA GTC Conference taking place from March 20-23. Our participation in this premier event highlights our leading role as an AI performance optimization solution.
The GTC is a must-attend event for anyone who wants to stay up-to-date on the latest trends and advancements in GPU technology and its applications. It will be held virtually and attendees can participate in the conference from anywhere with an internet connection.
Our tutorial and talk have been scheduled for the following times:
- Colossal-AI tutorial: March 20th, 12:00 am to 2:00 pm PDT
- Colossal-AI talk: March 22nd, 5:00 pm to 5:50 pm PDT
Both sessions will be presented by Prof. James Demmel, our Chief Strategy Officer and Distinguished Professor at UC Berkeley, as well as Prof. Yang You, our Founder, President and Chairman and Presidential Yong Professor at National University of Singapore. They will share how our open-source Colossal-AI system scales big models via optimization methods such as data, pipeline, tensor, sequence and auto parallelism, as well as heterogeneous system methods such as a zero-redundancy optimizer and offloading.
|
Prof. James Demmel |
Prof. Yang You |
Those interested in trying out our system prior to the tutorial or talk can visit our Colossal-AI GitHub repository to access the latest version. Feel free to reach out on our Colossal-AI developers Slack channel if you have any technical questions or cooperation intentions.
After the presentations, the slides will be made available for download on our website. The link to the download page will be provided during the presentations and shared on our social media channels (Twitter, LinkedIn, Facebook) for those who were unable to attend or would like to review the material again.
We extend our gratitude to NVIDIA for providing us with this incredible opportunity, plus we hope to see you at the GTC and look forward to discussing our insights and experiences with the broader community.
Comments