Efficient Large-Scale AI Workshop | Session 2: Training and inference efficiency
20This workshop was part of the Microsoft Research Summit 2022: https://www.microsoft.com/en-us/research/event/microsoft-research-summit-2022/
To bring AI to more people, models need to be cheaper to train and run, in terms of both computational and human resources. Thus, we will focus on increasing efficiency across various parts of the training and inference pipeline.
Learn more about the Efficient Large-Scale AI Workshop: https://www.microsoft.com/en-us/research/event/efficient-large-scale-ai-workshop/
0:00 Efficient Vision Transformer
Song Han, Massachusetts Institute of Technology
38:30 Large Scale MoE Models into Cloud Scale Production with Highly Efficient Inference and Training
Young Jin Kim, Microsoft Translator
Hany Awadalla, Azure AI Cognitive Services
1:43:32 LiteTransformerSearch: Training-free On-device Search for Efficient Autoregressive Language Models
Mojan Javaheripi, Microsoft Research Redmond and University of California San Diego